Product recommendation method and device, equipment, medium and program product
By introducing federated user profiles into existing product recommendation methods and combining user behavior and attribute data, the problem of singular financial product recommendations in existing technologies is solved, and more comprehensive and reliable product recommendations are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-11
- Publication Date
- 2026-04-10
AI Technical Summary
Existing product recommendation methods rely solely on users' own behavioral data, resulting in a limited range of financial product types and reduced reliability of recommendations.
By introducing federated user profiles, combining user behavior data, user attribute data, and federated learning results from different clients, user profile information is constructed, thereby filtering target recommended products from a group perspective.
This improves the comprehensiveness and reliability of product recommendations, ensuring the accuracy and diversity of product selection.
Smart Images
Figure CN121836918A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a product recommendation method, apparatus, device, medium, and program product. Background Technology
[0002] With the continuous development of the financial sector, product recommendation methods have emerged to better recommend financial products to users. Existing product recommendation methods generally recommend financial products based on users' historical behavioral data.
[0003] However, since the above technologies only involve users' own behavioral data, the types of financial products recommended are relatively limited, which reduces the reliability of product recommendations. Summary of the Invention
[0004] Therefore, it is necessary to provide a product recommendation method, apparatus, equipment, medium, and program product that can improve the reliability of product recommendation in order to address the above-mentioned technical problems.
[0005] Firstly, this application provides a product recommendation method, including:
[0006] In response to a product recommendation request for a target user, user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0007] Based on the federal user profile and user profile information, select the target recommended product from the candidate recommended products and feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user;
[0008] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0009] In one embodiment, the user profile information includes a user attribute profile and a user interest profile; the user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client, including:
[0010] Based on the target users' user behavior data, construct user interest profiles; based on the target users' user attribute data and the federated user profiles, construct user attribute profiles for the target users.
[0011] In one embodiment, selecting a target recommended product from candidate recommended products based on the federal user profile and user profile information includes:
[0012] Based on the user interest profile in the user profile information, a first recommendation list is determined from the candidate recommended products; based on the product purchase information of the first reference user associated with the federated user profile, a second recommendation list is determined from the candidate recommended products; based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information, a third recommendation list is determined from the candidate recommended products; based on the preset recommendation weights, the first recommendation list, the second recommendation list, and the third recommendation list are merged to obtain the target recommended product.
[0013] In one embodiment, determining a first recommendation list from candidate recommended products based on user interest profiles in user profile information includes:
[0014] Based on user interest profiles, determine the target user's interest weight for each product type, as well as the target user's level of interest in each candidate recommended product; based on the number of recommendable products and the interest weight for each product type, determine the number of product recommendations for each product type; based on the number of product recommendations for each product type and the target user's level of interest in each candidate recommended product, determine the first recommendation list from the candidate recommended products.
[0015] In one embodiment, a third recommendation list is determined from candidate recommended products based on user behavior data of a second reference user associated with the user attribute profile in the user profile information, including:
[0016] Based on the product purchase information of the first reference user associated with the federal user profile, a first candidate product list is selected from the candidate recommended products; based on the target user's purchased products and the first candidate product list, a second recommendation list is determined.
[0017] In one embodiment, the different clients also include other clients, and the method further includes:
[0018] Local user features are extracted from local user information in the target client. These local user features are then used to train a local feature model, yielding local model parameters. These local model parameters are sent to a central server, which instructs the central server to determine global model parameters based on the local model parameters and other model parameters sent by other clients. The global feature model parameters are then sent to the target client and other clients. These other model parameters are obtained by other clients training other models based on other user features, and the other feature models have the same model structure as the local feature model. The global model parameters are used to update the local feature model, and based on the updated local feature model, the local user information is processed to obtain a federated user profile associated with the target client.
[0019] Secondly, this application also provides a product recommendation device, comprising:
[0020] The information determination module is used to respond to product recommendation requests for target users by determining user profile information based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0021] The product determination module is used to select the target recommended product from the candidate recommended products based on the federated user profile and user profile information, and to feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user;
[0022] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0023] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:
[0024] In response to a product recommendation request for a target user, user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0025] Based on the federal user profile and user profile information, select the target recommended product from the candidate recommended products and feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user;
[0026] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0027] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:
[0028] In response to a product recommendation request for a target user, user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0029] Based on the federal user profile and user profile information, select the target recommended product from the candidate recommended products and feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user;
[0030] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0031] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:
[0032] In response to a product recommendation request for a target user, user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0033] Based on the federal user profile and user profile information, select the target recommended product from the candidate recommended products and feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user;
[0034] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0035] The aforementioned product recommendation methods, devices, equipment, media, and programs introduce a federated user profile. This profile is determined by analyzing the target user's behavior data, user attribute data, and the federated user profile associated with the target client. Based on the federated user profile and the user profile information, the target recommended product is selected from candidate recommended products. Compared to related technologies that determine the target recommended product solely based on user behavior data, this method, by introducing a federated user profile and user attribute data, can filter target recommended products for users from a group perspective, thereby ensuring the comprehensiveness of product selection and improving the reliability of product recommendations. Attached Figure Description
[0036] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0037] Figure 1 This is a flowchart illustrating a product recommendation method in one embodiment;
[0038] Figure 2 This is a schematic diagram of the process for determining user profile information in one embodiment;
[0039] Figure 3 This is a schematic diagram of the process for determining the target recommended product in one embodiment;
[0040] Figure 4 This is a schematic diagram of the process for determining the first recommendation list in one embodiment;
[0041] Figure 5 This is a flowchart illustrating the process of determining a second recommendation list in one embodiment;
[0042] Figure 6 This is a schematic diagram of the process for determining a federal user profile in one embodiment;
[0043] Figure 7 This is a flowchart illustrating the product recommendation method in another embodiment;
[0044] Figure 8 This is a structural block diagram of a product recommendation device in one embodiment;
[0045] Figure 9 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation
[0046] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0047] With the continuous development of the financial sector, product recommendation methods have emerged to better recommend financial products to users. Existing product recommendation methods generally recommend financial products based on users' historical behavioral data.
[0048] However, since the above technologies only involve users' own behavioral data, the types of financial products recommended are relatively limited, which reduces the reliability of product recommendations.
[0049] Based on this, in an exemplary embodiment, a product recommendation method is provided, taking the application of this method to clients associated with a central server as an example for illustration. The central server is a server that performs federated processing of model parameters uploaded by each client. Figure 1 As shown, the specific steps include:
[0050] S101, in response to a product recommendation request for a target user, determines user profile information based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0051] In this context, "target users" refers to registered users of applications with product recommendation features. A "product recommendation request" is used to initiate a product recommendation operation. "User behavior data" refers to the behavioral data generated when users interact with the target client. "User attribute data" refers to user attributes related to the target user, such as occupation, age, and location.
[0052] The target client is the client where the target user resides. The so-called federated user profile is used to characterize the overall features of users within the target client. Furthermore, the federated user profile is obtained by processing local user information within the target client, based on a local feature model of the target client obtained through federated learning across different clients; these different clients include the target client.
[0053] User profile information refers to information related to the target user's profile. Furthermore, user profile information includes user attribute profiles and user interest profiles. User attribute profiles are user profiles based on user attributes; user interest profiles are user profiles based on user interests.
[0054] Understandably, in order to better characterize the group interest features in the target client, federated learning can be performed based on the user information in each client, and the local user information in the target client can be processed based on the federated learning results to obtain a federated user profile in the target client.
[0055] Optionally, users can send a product recommendation request to the server by triggering a product recommendation control in the target client or by triggering a refresh control. After detecting the product recommendation request, user behavior data and user attribute data of the target user can be obtained based on the user identification information. Then, the user behavior data and user attribute data of the target user, as well as the federated user profile associated with the target client, are analyzed to obtain the user profile information of the target user.
[0056] For example, user behavior data, user attribute data, and the federated user profile associated with the target client can be input into a trained profile generation model, which will then output user profile information based on the user behavior data, user attribute data, and federated user profile.
[0057] It is worth noting that by introducing a federated user profile, data features of user information from other clients can be combined, making the federated user profile more comprehensive and richer in dimensions, thus avoiding the phenomenon of data silos.
[0058] S102, based on the federal user profile and user profile information, select the target recommended product from the candidate recommended products, and feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user.
[0059] The so-called candidate recommended products are the financial products that can be recommended to the target client. The so-called target recommended products are the financial products that need to be recommended to the target users.
[0060] Optionally, the types of products that users are interested in and the degree of interest can be determined based on the federated user profile, user attribute profile, and user interest profile; then, the target recommended product can be selected from the candidate recommended products based on the types of products that users are interested in and the degree of interest.
[0061] Furthermore, the target recommended products can be sent to the target client and displayed through the product display page associated with the target client, thereby recommending the target recommended products to the target users.
[0062] The product recommendation method described above introduces a federated user profile. This profile is determined by analyzing the target user's behavior data, user attribute data, and the federated user profile associated with the target client. Based on the federated user profile and the user profile information, the target recommended product is selected from the candidate recommended products. Compared to related technologies that determine the target recommended product solely based on user behavior data, this method, by introducing a federated user profile and user attribute data, can filter target recommended products for users from a group perspective, thereby ensuring the comprehensiveness of product selection and improving the reliability of product recommendations.
[0063] Based on the above embodiments, in this application embodiment, the user profile information includes user attribute profiles and user interest profiles. Furthermore, an optional method for determining user profile information is provided, such as... Figure 2 As shown, the specific steps include:
[0064] S201, Construct user interest profiles based on the user behavior data of the target users.
[0065] User behavior data may include, but is not limited to, click data, browsing data, purchase record data, and rating data related to financial products.
[0066] Optionally, dynamic user interest features can be extracted from user behavior data based on a sliding time window (such as the past 30 days). These dynamic user interest features may include, but are not limited to, accessed product categories, purchase frequency, product ratings, and interest decay coefficients.
[0067] Furthermore, dynamic user interest features can be input into a trained user interest model, which then outputs a user interest profile based on these features. This user interest profile can be updated in real time according to changes in user behavior data.
[0068] S202, construct the target user's user attribute profile based on the target user's user attribute data and the federated user profile.
[0069] User attribute data may include, but is not limited to, the target user's age, gender, occupation, geographical location, income, risk tolerance rating, etc.
[0070] Optionally, an initial attribute profile of the target user can be constructed based on the user attribute data of the target user; then, a federated user profile can be used to adjust the initial attribute profile to obtain the user attribute profile of the target user.
[0071] Alternatively, user attribute features can be extracted from user attribute data based on preset attribute correlation rules. These attribute correlation rules are the association rules between attributes in each dimension and financial products; for example, "high income characteristics + high risk tolerance" corresponds to aggressive investment attributes.
[0072] Furthermore, based on the distribution of group attributes in the federal user profile, the weights of attribute features in the user attribute features can be adjusted to construct a user attribute profile. For example, if "users aged 30 and above" account for more than 60% in the federal user profile, the weight of the "age" feature in the user attribute features can be increased.
[0073] In this embodiment of the application, by constructing user interest profiles based on user behavior data and user attribute profiles based on user attribute data and federated user profiles, the accuracy of the user profile information can be guaranteed.
[0074] Based on the above embodiments, this application provides an optional method for determining target recommended products, such as... Figure 3 As shown, the specific steps include:
[0075] S301, Based on the user interest profile in the user profile information, determine the first recommendation list from the candidate recommended products.
[0076] The so-called first recommendation list is a product recommendation list based on user interests, and the products in the first recommendation list are sorted from high to low based on the user's level of interest.
[0077] Optionally, by analyzing user interest profiles, the target user's level of interest in each product type can be determined; then, the candidate recommended products can be sorted from high to low according to the target user's level of interest in each product type to obtain the first recommendation list.
[0078] S302, Based on the product purchase information of the first reference user associated with the federal user profile, determine the second recommendation list from the candidate recommended products.
[0079] The so-called second recommendation list is a product recommendation list based on the group dimension, and the products in the second recommendation list are sorted from high to low based on the local group's interest level. The so-called first reference user is the local user in the target client that matches the user characteristics in the federated user profile. The so-called product purchase information is the information related to the financial products purchased by the first reference user.
[0080] Optionally, by analyzing the federated user profile, a first reference user can be selected from the local users associated with the target client. Based on the first reference user's product purchase information, the purchase frequency of each type of product by the first reference user can be determined. Then, the candidate recommended products can be sorted from high to low according to the first reference user's purchase frequency for each type of product to obtain a second recommendation list.
[0081] S303, Based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information, determine the third recommendation list from the candidate recommended products.
[0082] The so-called third recommendation list is a product recommendation list based on user attributes, and the products in the third recommendation list are sorted from high to low according to the level of interest of the user's attribute group. The so-called second reference users are users who match the user attributes.
[0083] Optionally, a second reference user can be selected from local users based on user attribute profiles. For example, users with attribute feature similarity ≥ 90% can be used as the second reference user. Then, a third recommendation list can be selected from the candidate recommended products based on the user behavior data of the second reference user.
[0084] For example, the top 3 products purchased by the second reference user can be used as reference recommended products, and a third recommendation list can be generated in descending order of purchase frequency.
[0085] S304. Based on the preset recommendation weights, the first recommendation list, the second recommendation list, and the third recommendation list are merged to obtain the target recommended product.
[0086] The so-called preset recommendation weights are the weight values corresponding to the first, second, and third recommendation lists, determined according to their importance. For example, the preset recommendation weight for the first recommendation list is 0.5, the preset recommendation weight for the second recommendation list is 0.3, and the preset recommendation weight for the third recommendation list is 0.2.
[0087] Optionally, a preset recommendation weight can be used to weight the recommendation status of each candidate product in the first recommendation list, the second recommendation list, and the third recommendation list. Then, the weighted first recommendation list, the second recommendation list, and the third recommendation list can be merged to obtain the target recommended products with a preset number.
[0088] For example, the initial recommendation value of each candidate product in each recommendation list can be determined based on the arrangement position of each candidate product in the first recommendation list, the second recommendation list, and the third recommendation list. Then, the initial recommendation value of each candidate product can be weighted using a preset recommendation weight, and the weighted recommendation values of each candidate product in each recommendation list can be merged to obtain the final recommendation value of each candidate product.
[0089] Furthermore, based on the final recommendation value of each candidate product, the candidate products can be ranked, and a preset number of the top-ranked candidate products can be used as the target recommended products.
[0090] In this embodiment of the application, by determining a product recommendation list from each profile and combining the product recommendation lists to determine the target recommended product, the accuracy of the target recommended product determination can be guaranteed.
[0091] Based on the above embodiments, this application provides an optional method for determining the first recommendation list, such as... Figure 4 As shown, the specific steps include:
[0092] S401, based on user interest profiles, determine the target user's interest weight for each product type, as well as the target user's level of interest in each candidate recommended product.
[0093] Among them, the so-called interest weight is used to characterize the degree of user interest in various types of products.
[0094] Optionally, the interest weight of the target user for each product type can be determined based on the user interest characteristics represented in the user interest profile, and the degree of interest of the target user for each candidate recommended product can be determined based on the user behavior characteristics in the user interest profile.
[0095] S402, determine the number of recommended products for each product type based on the number of recommended products and the interest weight of each product type.
[0096] The "number of recommendable products" refers to the number of products that can be filtered based on user interests; that is, the total number of products that can be displayed in the first recommendation list. The "number of recommended products" refers to the number of products selected for recommendation within each product type.
[0097] Optionally, the number of recommended products can be weighted by the interest weight of each product type to obtain the number of recommended products for each product type.
[0098] For example, if the interest weight for product type A is 0.6, the interest weight for product type B is 0.3, and the interest weight for product type C is 0.1, and the number of recommended products is 10, then the number of recommended products for product type A is 6, the number of recommended products for product type B is 3, and the number of recommended products for product type C is 1.
[0099] S403. Based on the number of product recommendations for each product type and the target user's interest in each candidate recommended product, determine the first recommendation list from the candidate recommended products.
[0100] Optionally, for each product type, the candidate recommended products are sorted in descending order of interest based on their level of interest in that product type, and the top-ranked candidate recommended products are selected as the products to be recommended.
[0101] Furthermore, the products to be recommended for each product type are sorted in descending order of interest level to generate the first recommendation list.
[0102] In this embodiment of the application, the first recommendation list is determined from the candidate recommended products by using the target user's interest weight for each product type and the target user's interest in each candidate recommended product, which can ensure the reliability of the determination of the first recommendation list.
[0103] Based on the above embodiments, this application provides an optional method for determining the second recommendation list, such as... Figure 5 As shown, the specific steps include:
[0104] S501, based on the product purchase information of the first reference user associated with the federal user profile, select the first candidate product list from the candidate recommended products.
[0105] The so-called first candidate product list is a product list built based on the number of times local users have purchased each candidate recommended product.
[0106] Optionally, based on the product purchase information of the first reference user, the candidate recommended products can be sorted in descending order of purchase frequency, and a first candidate product list can be constructed based on the first preset number of candidate recommended products.
[0107] S502, determine the second recommendation list based on the target user's purchased products and the first candidate product list.
[0108] The so-called "purchased products" refer to the candidate recommended products that the target user has purchased.
[0109] Optionally, products already purchased by the target user can be removed from the first candidate product list to obtain a second recommendation list.
[0110] In this embodiment of the application, by determining the second recommendation list based on the product purchase information of the first reference user and the products already purchased by the target user, a product recommendation list that the target user has not purchased can be constructed, thereby ensuring the reliability of the second recommendation list.
[0111] Based on the above embodiments, in this application embodiment, different clients also include other clients, providing an optional method for determining federal user profiles, such as... Figure 6 As shown, the specific steps include:
[0112] S601 extracts local user features from local user information in the target client.
[0113] The so-called local user information refers to the user information of each user in the target client, which may include user behavior data. The so-called local user characteristics refer to the user characteristics of each user in the target client.
[0114] Optionally, the local user information in the target client can be preprocessed (data cleaning, feature normalization) to extract local user features (such as local user preferred product categories, average purchase cycle, etc.).
[0115] S602 uses local user features to train a local feature model, thereby obtaining the local model parameters of the local feature model.
[0116] The so-called local feature model refers to the group feature model in the target client. The so-called local model parameters refer to the model parameters of the local feature model.
[0117] Optionally, local user features can be used to train the local feature model to obtain the local model parameters. For example, local user features can be used to train the local feature model and the gradient of the model parameters can be calculated.
[0118] It is worth noting that, in order to ensure the reliability of the local feature model, a global feature model (such as a feature extraction model based on logistic regression or a lightweight neural network) can be initialized through a central server, and the model parameters can be distributed to each client to build the feature model in each client.
[0119] S603 sends the local model parameters to the central server to instruct the central server to determine the global model parameters based on the local model parameters and other model parameters sent by other clients, and sends the global feature model parameters to the target client and other clients.
[0120] Among them, the other model parameters are obtained by other clients training other models based on other user features, and the model structure of other feature models is the same as that of local feature models. The global model parameters are the model parameters after federated processing.
[0121] Optionally, local model parameters from the target client, as well as other model parameters from other clients, can be uploaded to the central server. After receiving the local model parameters and other model parameters, the central server can use algorithms such as FedAvg and FedMomentum to perform weighted aggregation of the local model parameters and other model parameters (the weights are positively correlated with the number of users on the client) to obtain the global model parameters.
[0122] It is worth noting that if the aggregated model loss function (such as MSE, cross-entropy) does not converge, the central server can first distribute the updated global model parameters to each client and repeat the "local training - parameter upload - aggregation update" steps (the number of iterations is preset to 10-50 times, or until the loss function converges to the threshold) until the model loss function converges.
[0123] For example, the central server uses a federated averaging algorithm to aggregate the gradients of 30 clients (weights are allocated according to the number of users on each client, such as client X having a user share of 15% and a weight of 0.15) and updates the global model; after 20 iterations, the global model loss function (cross-entropy) converges to below 0.05, and training stops.
[0124] S604 uses global model parameters to update the local feature model, and processes local user information based on the updated local feature model to obtain the federated user profile associated with the target client.
[0125] Optionally, the local feature model can be updated based on the global model parameters to obtain the federated local feature model. Then, local user information can be input into the federated local feature model, which will output the federated user profile of the target client based on the local user information and model parameters.
[0126] It is worth noting that, in order to ensure the reliability of the federated user profile, the model gradient of each client can be controlled to re-upload the newly added user behavior data locally according to the preset parameter collection cycle; the central server aggregates and updates the global model parameters, and each client synchronously updates its local federated user profile. That is, the federated user profile of the target user is adjusted in real time with the update of the global model parameters to ensure the timeliness of the recommendation.
[0127] For example, it can be set to trigger a federated model update at 3:00 AM every day. That is, at 3:00 AM every day, each client uploads the gradient of newly added user behavior data for that day; the central server aggregates and updates the global model and instructs each client to update the federated user profile in real time.
[0128] In this embodiment of the application, by performing federated processing on the user information in each client, a federated user profile for each client is obtained, which can ensure the reliability of the determination of the federated user profile.
[0129] Figure 7 This is a flowchart illustrating a product recommendation method in another embodiment. Based on the above embodiments, this embodiment provides an optional example of a product recommendation method. (Combined with...) Figure 7 The specific implementation process is as follows:
[0130] S701 responds to product recommendation requests for target users by constructing user interest profiles based on user behavior data of the target users.
[0131] S702, constructs a user attribute profile of the target user based on the target user's user attribute data and the federated user profile.
[0132] Among them, the federated user profile is a local feature model of the target client obtained by federated learning on different clients, which is obtained by processing the local user information in the target client; the different clients include the target client and other clients.
[0133] Optionally, local user features from the target client are extracted from local user information; the local feature model is trained using the local user features to obtain local model parameters; the local model parameters are sent to the central server to instruct the central server to determine global model parameters based on the local model parameters and other model parameters sent by other clients, and then send the global feature model parameters to the target client and other clients; wherein, the other model parameters are obtained by other clients training other models based on other user features, and the model structures of the other feature models and the local feature models are the same; the local feature model is updated using the global model parameters, and the local user information is processed based on the updated local feature model to obtain the federated user profile associated with the target client.
[0134] S703, based on the user interest profile in the user profile information, determines the first recommended list from the candidate recommended products.
[0135] Optionally, based on user interest profiles, determine the target user's interest weight for each product type and the target user's interest level for each candidate recommended product; based on the number of recommendable products and the interest weight for each product type, determine the number of product recommendations for each product type; based on the number of product recommendations for each product type and the target user's interest level for each candidate recommended product, determine the first recommendation list from the candidate recommended products.
[0136] S704, based on the product purchase information of the first reference user associated with the federal user profile, determine the second recommendation list from the candidate recommended products.
[0137] Optionally, a first candidate product list is selected from the candidate recommended products based on the product purchase information of the first reference user associated with the federal user profile; a second recommendation list is determined based on the target user's purchased products and the first candidate product list.
[0138] S705, based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information, determine the third recommendation list from the candidate recommended products.
[0139] S706, according to the preset recommendation weights, merge the first recommendation list, the second recommendation list and the third recommendation list to obtain the target recommended product.
[0140] S707 feeds back the target recommended products to the target client so that the target client can display the target recommended products to the target user.
[0141] The specific processes of S701-S707 described above can be found in the description of the above method embodiments. Their implementation principles and technical effects are similar, and will not be repeated here.
[0142] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.
[0143] Based on the same inventive concept, this application also provides a product recommendation apparatus for implementing the product recommendation method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more product recommendation apparatus embodiments provided below can be found in the limitations of the product recommendation method described above, and will not be repeated here.
[0144] In one exemplary embodiment, such as Figure 8 As shown, a product recommendation device 1 is provided, including: an information determination module 10 and a product determination module 20, wherein:
[0145] The information determination module 10 is used to respond to a product recommendation request for a target user by determining user profile information based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client.
[0146] Product determination module 20 is used to select target recommended products from candidate recommended products based on the federated user profile and user profile information, and to feed back the target recommended products to the target client so that the target client can display the target recommended products to the target user;
[0147] Among them, the federated user profile is a local feature model of the target client obtained by federated learning of different clients, which is obtained by processing the local user information in the target client; the different clients include the target client.
[0148] In an exemplary embodiment, the user profile information includes a user attribute profile and a user interest profile; the information determination module 10 is specifically used for:
[0149] Based on the target users' user behavior data, construct user interest profiles; based on the target users' user attribute data and the federated user profiles, construct user attribute profiles for the target users.
[0150] In one exemplary embodiment, the product determination module 20 is specifically used for:
[0151] Based on the user interest profile in the user profile information, a first recommendation list is determined from the candidate recommended products; based on the product purchase information of the first reference user associated with the federated user profile, a second recommendation list is determined from the candidate recommended products; based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information, a third recommendation list is determined from the candidate recommended products; based on the preset recommendation weights, the first recommendation list, the second recommendation list, and the third recommendation list are merged to obtain the target recommended product.
[0152] In one exemplary embodiment, the product determination module 20 is further configured to:
[0153] Based on user interest profiles, determine the target user's interest weight for each product type, as well as the target user's level of interest in each candidate recommended product; based on the number of recommendable products and the interest weight for each product type, determine the number of product recommendations for each product type; based on the number of product recommendations for each product type and the target user's level of interest in each candidate recommended product, determine the first recommendation list from the candidate recommended products.
[0154] In one exemplary embodiment, the product determination module 20 is further configured to:
[0155] Based on the product purchase information of the first reference user associated with the federal user profile, a first candidate product list is selected from the candidate recommended products; based on the target user's purchased products and the first candidate product list, a second recommendation list is determined.
[0156] In one exemplary embodiment, the different clients also include other clients, and the product recommendation device 1 further includes a federated learning module, wherein the federated learning module is specifically used for:
[0157] Local user features are extracted from local user information in the target client. These local user features are then used to train a local feature model, yielding local model parameters. These local model parameters are sent to a central server, which instructs the central server to determine global model parameters based on the local model parameters and other model parameters sent by other clients. The global feature model parameters are then sent to the target client and other clients. These other model parameters are obtained by other clients training other models based on other user features, and the other feature models have the same model structure as the local feature model. The global model parameters are used to update the local feature model, and based on the updated local feature model, the local user information is processed to obtain a federated user profile associated with the target client.
[0158] Each module in the aforementioned recommended product device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.
[0159] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 9As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a product recommendation method. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.
[0160] Those skilled in the art will understand that Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.
[0161] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.
[0162] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.
[0163] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.
[0164] It should be noted that the data involved in this application (including but not limited to user behavior data) is all data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.
[0165] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.
[0166] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.
[0167] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A product recommendation method, characterized in that, The method includes: In response to a product recommendation request for a target user, user profile information is determined based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client. Based on the federated user profile and the user profile information, a target recommended product is selected from the candidate recommended products, and the target recommended product is fed back to the target client so that the target client can display the target recommended product to the target user; The federated user profile is obtained by processing local user information in the target client based on the local feature model of the target client obtained through federated learning of different clients; the different clients include the target client.
2. The method according to claim 1, characterized in that, The user profile information includes user attribute profiles and user interest profiles; determining the user profile information based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client includes: Based on the user behavior data of the target users, construct user interest profiles; Based on the target user's user attribute data and the federated user profile, construct the target user's user attribute profile.
3. The method according to claim 1, characterized in that, The step of selecting a target recommended product from candidate recommended products based on the federal user profile and the user profile information includes: Based on the user interest profile in the user profile information, a first recommendation list is determined from the candidate recommended products; Based on the product purchase information of the first reference user associated with the federal user profile, a second recommendation list is determined from the candidate recommended products; Based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information, a third recommendation list is determined from the candidate recommended products; Based on preset recommendation weights, the first recommendation list, the second recommendation list, and the third recommendation list are merged to obtain the target recommended product.
4. The method according to claim 3, characterized in that, The step of determining a first recommendation list from candidate recommended products based on the user interest profile in the user profile information includes: Based on the user interest profile, determine the target user's interest weight for each product type, and the target user's degree of interest in each candidate recommended product; The number of products to be recommended for each product type is determined based on the number of products that can be recommended and the interest weight of each product type. Based on the number of product recommendations for each product type and the target user's level of interest in each candidate recommended product, a first recommendation list is determined from the candidate recommended products.
5. The method according to claim 3, characterized in that, The step of determining a third recommendation list from the candidate recommended products based on the user behavior data of the second reference user associated with the user attribute profile in the user profile information includes: Based on the product purchase information of the first reference user associated with the federal user profile, select the first candidate product list from the candidate recommended products; A second recommendation list is determined based on the target user's purchased products and the first candidate product list.
6. The method according to claim 1, characterized in that, The different clients also include other clients, and the method further includes: Extract local user features from the local user information in the target client; Using the local user features, the local feature model is trained to obtain the local model parameters of the local feature model; The local model parameters are sent to the central server to instruct the central server to determine the global model parameters based on the local model parameters and other model parameters sent by other clients, and to send the global feature model parameters to the target client and the other clients; wherein, the other model parameters are obtained by the other clients by training other models based on other user features, and the other feature models and the local feature models have the same model structure; The local feature model is updated using the global model parameters, and the local user information is processed based on the updated local feature model to obtain the federal user profile associated with the target client.
7. A product recommendation device, characterized in that, The device includes: The information determination module is used to respond to a product recommendation request for a target user by determining user profile information based on the target user's user behavior data, user attribute data, and the federated user profile associated with the target user's target client. The product determination module is used to select a target recommended product from the candidate recommended products based on the federated user profile and the user profile information, and to feed back the target recommended product to the target client so that the target client can display the target recommended product to the target user; The federated user profile is obtained by processing local user information in the target client based on the local feature model of the target client obtained through federated learning of different clients; the different clients include the target client.
8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.