Training method of fusion recommendation model, item recommendation method and user recommendation method

Abstract

The application relates to a training method of a fusion recommendation model, an item recommendation method and a user recommendation method, and relates to the technical field of automobile big data application. The method comprises the following steps: determining the interest feature of a preset user, and determining the difference feature of the preset user based on the user feature of the preset user and the user feature of each associated user in the plurality of associated users; inputting the interest feature of the preset user and the difference feature of the preset user into an initial item recommendation model to obtain a first prediction result of the preset user, and inputting the user feature of the preset user and the user feature of each user into an initial user recommendation model to obtain a second prediction feature of the preset user; and training the initial fusion recommendation model based on the first prediction result of the preset user and the second prediction result of the preset user, so as to generate a target fusion recommendation model. Therefore, the effectiveness of item recommendation or user recommendation can be improved.

Description

Technical Field

[0001] This application relates to the field of automotive big data application technology, specifically to a training method for a fusion recommendation model, an item recommendation method, and a user recommendation method. Background Technology

[0002] Currently, based on statistical and collaborative filtering methods, an item that a user is interested in can be identified and recommended to that user.

[0003] However, due to the rapid growth of internet data and the diversification of user behavior, the above methods may not be able to accurately identify items that users are interested in, thus affecting the effectiveness of item recommendations. Summary of the Invention

[0004] This application provides a training method for a recommendation model, an item recommendation method, and a user recommendation method, to at least address the technical problem in related technologies where the rapid growth of internet data and the diversification of user behavior mean that methods based on statistics and collaborative filtering may not accurately identify items of interest to users, thus affecting the effectiveness of item recommendations. The technical solution of this application is as follows:

[0005] According to a first aspect of this application, a training method for a fusion recommendation model is provided, comprising: determining user characteristics of a preset user, item characteristics of each of a plurality of items, and user characteristics of each of a plurality of users, wherein the plurality of items are items of interest to the preset user, and the plurality of users include a plurality of associated users, which are users who have interactive behaviors with the preset user; determining interest characteristics of the preset user based on the user characteristics of the preset user and the item characteristics of each item, and determining difference characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each of the plurality of associated users, wherein the interest characteristics of the preset user are used to characterize the degree of interest of the preset user in each item, and the difference characteristics of the preset user are used to characterize the degree of difference between the preset user and each of the associated users; and then... The interest features and differential features of a preset user are input into an initial item recommendation model to obtain a first prediction result for the preset user. The user features of the preset user and the user features of each user are then input into an initial user recommendation model to obtain a second prediction feature for the preset user. The initial item recommendation model is an item recommendation model included in the initial fusion recommendation model, and the initial user recommendation model is a user recommendation model included in the initial fusion recommendation model. Based on the first and second prediction results of the preset user, the initial fusion recommendation model is trained to generate a target fusion recommendation model. The target fusion recommendation model includes a target item recommendation model and a target user recommendation model. The target item recommendation model is used to recommend items to a user, and the target user recommendation model is used to recommend other users to a user.

[0006] Based on the aforementioned technical means, the electronic device can determine the user's interest characteristics based on the user characteristics of the preset user and the item characteristics of each item among multiple items. Furthermore, based on the user characteristics of the preset user and the user characteristics of each associated user among multiple associated users, the electronic device can determine the user's difference characteristics. The electronic device can then train an initial item recommendation model based on the user's interest characteristics and difference characteristics to generate an item recommendation model with high prediction accuracy (i.e., a target item recommendation model). Additionally, the electronic device can train an initial user recommendation model based on the user characteristics of the preset user and the user characteristics of each user among multiple users to generate a user recommendation model with high prediction accuracy (i.e., a target user recommendation model). This leads to a fusion recommendation model with high prediction accuracy (i.e., a target fusion recommendation model). The electronic device can then perform item and user recommendations based on this target fusion recommendation model, thereby improving the effectiveness of item and user recommendations.

[0007] In one possible implementation, training the initial fusion recommendation model based on the first prediction result and the second prediction result of the preset user to generate a target fusion recommendation model includes: obtaining the first true result and the second true result of the preset user; determining a first loss based on the first true result and the first prediction result of the preset user, and determining a second loss based on the second true result and the second prediction result of the preset user, wherein the first loss is used to characterize the degree of inconsistency between the first true result and the first prediction result of the preset user, and the second loss is used to characterize the degree of inconsistency between the second true result and the second prediction result of the preset user; determining a target loss based on the first loss and the second loss; and updating the parameters in the initial fusion recommendation model based on the target loss to obtain the target fusion recommendation model.

[0008] Based on the aforementioned technical means, since the first predicted result for the preset user is the item with high interest predicted by the initial item recommendation model, and the first true result for the preset user is the actual item the preset user is interested in, the first loss determined by the electronic device based on the first predicted result and the first true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Furthermore, since the second predicted result for the preset user is other users with high relevance predicted by the initial user recommendation model, and the second true result for the preset user is the actual relevance of the preset user, the second loss determined by the electronic device based on the second predicted result and the second true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Then, based on the first loss and the second loss, the electronic device can obtain the target loss, and then update the parameters in the initial fusion recommendation model based on the target loss, which can conveniently and quickly generate the target fusion recommendation model, improving the efficiency of model training.

[0009] In one possible implementation, the above-mentioned determination of the user characteristics of a preset user, the item characteristics of each of the multiple items, and the user characteristics of each of the multiple users includes: obtaining the user information of the preset user, the item identifier of each item, and the user information of each user, wherein the user information of the preset user includes one or more of the user name, the user age, and the location information of the preset user; and performing feature recognition on the user information of the preset user, the item identifier of each item, and the user information of each user respectively to obtain the user characteristics of the preset user, the item characteristics of each item, and the user characteristics of each user.

[0010] Based on the aforementioned technical means, electronic devices can perform feature recognition based on the user information of the preset user, the item identifier of each item, and the user information of each associated user, thereby accurately and effectively obtaining the user characteristics of the preset user, the item characteristics of each item, and the user characteristics of each associated user.

[0011] In one possible implementation, determining the differential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each of the plurality of associated users includes: determining the product of the user characteristics of the preset user and the user characteristics of each associated user as the potential characteristics of the preset user; determining the difference between the potential characteristics of the preset user and the user characteristics of the preset user as the association characteristics of the preset user; and determining the square of the association characteristics of the preset user as the differential characteristics of the preset user.

[0012] Based on the aforementioned technical means, the electronic device can determine the potential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the preset user, the potential characteristics of the preset user can be determined. Finally, based on the potential characteristics of the preset user, the differential characteristics of the preset user can be accurately and effectively determined, thereby accurately training the model.

[0013] According to a second aspect of this application, an item recommendation method is provided, comprising: determining user characteristics of a target user, item characteristics of each of at least one item, and user characteristics of each of at least one associated user, wherein the at least one item is an item of interest to the target user, and the at least one associated user is a user who has interacted with the target user; determining interest characteristics of the target user based on the user characteristics of the target user and the item characteristics of each item, and determining difference characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user, wherein the interest characteristics of the target user are used to characterize the degree of interest of the target user in each item, and the difference characteristics of the target user are used to characterize the degree of difference between the target user and each associated user; inputting the interest characteristics and the difference characteristics of the target user into a target item recommendation model to obtain a first prediction result for the target user, wherein the first prediction result for the target user includes the item identifier of the target item, the target item recommendation model is an item recommendation model included in a target fusion recommendation model, the target fusion recommendation model being trained based on the training method of the fusion recommendation model of the first aspect; and recommending the target item to the target user.

[0014] Based on the aforementioned technical means, this application can determine the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users. Based on the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users, the interest characteristics and the difference characteristics of the target user can be obtained. The electronic device inputs the interest characteristics and the difference characteristics of the target user into the target item recommendation model, thereby accurately and effectively obtaining the recommended items for the target user and recommending the target item to the target user.

[0015] In one possible implementation, determining the user characteristics of the target user, the item characteristics of each of at least one item, and the user characteristics of each of at least one associated user includes: obtaining the user information of the target user, the item identifier of each item, and the user information of each associated user, wherein the user information of the target user includes one or more of the user name, the user age, and the location information of the target user; and performing feature recognition on the user information of the target user, the item identifier of each item, and the user information of each associated user to obtain the user characteristics of the target user, the item characteristics of each item, and the user characteristics of each associated user.

[0016] Based on the aforementioned technical means, this application can perform feature recognition based on the obtained user information of the target user, the item identifier of each item, and the user information of each associated user, thereby accurately and effectively obtaining the user characteristics of the target user, the item characteristics of each item, and the user characteristics of each associated user.

[0017] In one possible implementation, determining the differential characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user includes: determining the product of the user characteristics of the target user and the user characteristics of each associated user as the potential characteristics of the target user; determining the difference between the potential characteristics of the target user and the user characteristics of the target user as the associated characteristics of the target user; and determining the square of the associated characteristics of the target user as the differential characteristics of the target user.

[0018] Based on the aforementioned technical means, this application can determine the potential characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the target user, the potential characteristics of the target user can be determined. Finally, based on the potential characteristics of the target user, the associated characteristics of the target user can be accurately and effectively determined, thereby accurately training the model.

[0019] According to a third aspect of this application, a user recommendation method is provided, comprising: determining user features of a target user and user features of each user among at least one user; inputting the user features of the target user and the user features of each user into a target user recommendation model to obtain a second prediction result for the target user, the second prediction result for the target user being used to characterize whether to recommend each user to the target user, the target user recommendation model being a user recommendation model included in a target fusion recommendation model; and recommending other users to the target user if the second prediction result for the target user characterizes recommending other users to the target user, the other users being users included among the at least one user.

[0020] Based on the aforementioned technical means, the electronic device first determines the user characteristics of the target user and the user characteristics of each of the at least one users. The electronic device then inputs the user characteristics of the target user and the user characteristics of each of the at least one users into the target user recommendation model, thereby accurately and effectively determining whether to recommend other users to the target user. Furthermore, if the second prediction result indicates that other users are recommended to the target user, it indicates that the other users are users of interest to the target user. In this case, the electronic device can accurately and effectively recommend other users to the target user.

[0021] According to the fourth aspect provided in this application, a training apparatus for a fusion recommendation model is provided, comprising: a determining unit, a processing unit, and a generating unit. The determining unit is configured to determine user characteristics of a preset user, item characteristics of each of a plurality of items, and user characteristics of each of the plurality of users, wherein the plurality of items are items of interest to the preset user, and the plurality of users include a plurality of associated users, which are users who have interacted with the preset user; the determining unit is configured to determine the preset user's interest characteristics based on the user characteristics of the preset user and the item characteristics of each item, and to determine the preset user's difference characteristics based on the user characteristics of the preset user and the user characteristics of each of the plurality of associated users, wherein the preset user's interest characteristics characterize the degree of interest of the preset user in each item, and the preset user's difference characteristics characterize the degree of difference between the preset user and each of the associated users; the processing unit is configured to process the preset user's interest characteristics... The features and differential features of the preset user are input into the initial item recommendation model to obtain the first prediction result of the preset user. The user features of the preset user and the user features of each user are input into the initial user recommendation model to obtain the second prediction result of the preset user. The initial item recommendation model is an item recommendation model included in the initial fusion recommendation model, and the initial user recommendation model is a user recommendation model included in the initial fusion recommendation model. The above-mentioned generation unit is used to train the initial fusion recommendation model based on the first prediction result and the second prediction result of the preset user to generate a target fusion recommendation model. The target fusion recommendation model includes a target item recommendation model and a target user recommendation model. The target item recommendation model is used to recommend items to a user, and the target user recommendation model is used to recommend other users to a user.

[0022] In one possible implementation, the training apparatus for the aforementioned fusion recommendation model further includes: an acquisition unit. The acquisition unit is configured to acquire a first true result and a second true result of the preset user; the determination unit is specifically configured to determine a first loss based on the first true result and a first prediction result of the preset user, and to determine a second loss based on the second true result and a second prediction result of the preset user, wherein the first loss characterizes the degree of inconsistency between the first true result and the first prediction result of the preset user, and the second loss characterizes the degree of inconsistency between the second true result and the second prediction result of the preset user; the determination unit is further specifically configured to determine a target loss based on the first loss and the second loss; and the generation unit is specifically configured to update the parameters in the initial fusion recommendation model based on the target loss to obtain the target fusion recommendation model.

[0023] In one possible implementation, the training apparatus for the aforementioned fusion recommendation model further includes: an acquisition unit. The acquisition unit acquires user information of the preset user, item identifiers of each item, and user information of each user; the user information of the preset user includes one or more of the following: the user name of the preset user, the user age of the preset user, and the location information of the preset user; the processing unit performs feature recognition on the user information of the preset user, the item identifiers of each item, and the user information of each user, respectively, to obtain user features of the preset user, item features of each item, and user features of each user.

[0024] In one possible implementation, the determining unit is further specifically configured to determine the product between the user characteristics of the preset user and the user characteristics of each associated user as the potential characteristics of the preset user; the determining unit is further specifically configured to determine the difference between the potential characteristics of the preset user and the user characteristics of the preset user as the association characteristics of the preset user; and the determining unit is further specifically configured to determine the square of the association characteristics of the preset user as the difference characteristics of the preset user.

[0025] According to a fifth aspect of this application, an item recommendation apparatus is provided, comprising: a determining unit and a processing unit. The determining unit is configured to determine user characteristics of a target user, item characteristics of each of at least one item, and user characteristics of each of at least one associated user, wherein the at least one item is an item of interest to the target user, and the at least one associated user is a user who has interacted with the target user; the determining unit is configured to determine the target user's interest characteristics based on the target user's user characteristics and the item characteristics of each item, and to determine the target user's difference characteristics based on the target user's user characteristics and the user characteristics of each associated user, wherein the target user's interest characteristics characterize the degree of interest the target user has in each item, and the target user's difference characteristics characterize the degree of difference between the target user and each associated user; the processing unit is configured to input the target user's interest characteristics and the target user's difference characteristics into a target item recommendation model to obtain a first prediction result for the target user, the first prediction result including the item identifier of the target item; and the processing unit is configured to recommend the target item to the target user.

[0026] In one possible implementation, the item recommendation device further includes an acquisition unit. The acquisition unit is configured to acquire user information of the target user, item identifiers of each item, and user information of each associated user. The user information of the target user includes one or more of the target user's username, user age, and location information. The processing unit is specifically configured to perform feature recognition on the user information of the target user, the item identifiers of each item, and the user information of each associated user, respectively, to obtain user characteristics of the target user, item characteristics of each item, and user characteristics of each associated user.

[0027] In one possible implementation, the determining unit is specifically configured to determine the potential feature of the target user as the product of the user features of the target user and the user features of each associated user; the determining unit is also specifically configured to determine the difference between the potential feature of the target user and the user features of the target user as the associated feature of the target user; the determining unit is also specifically configured to determine the square of the associated feature of the target user as the difference feature of the target user.

[0028] According to a sixth aspect of this application, a user recommendation apparatus is provided, comprising: a determining unit and a processing unit. The determining unit is configured to determine user characteristics of a target user and user characteristics of each of at least one user; the processing unit is configured to input the user characteristics of the target user and the user characteristics of each user into a target user recommendation model to obtain a second prediction result for the target user, the second prediction result representing whether to recommend each user to the target user, the target user recommendation model being a user recommendation model included in a target fusion recommendation model; the processing unit is configured to recommend other users to the target user if the second prediction result represents recommending other users to the target user, the other users being users included in the at least one user.

[0029] According to a seventh aspect provided in this application, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute instructions to implement the methods of the first aspect, the second aspect, the third aspect, and any possible implementation thereof.

[0030] According to the eighth aspect provided in this application, a computer-readable storage medium is provided that, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the methods of the first aspect, the second aspect, the third aspect, and any possible implementation thereof.

[0031] Therefore, the above-mentioned technical features of this application have the following beneficial effects:

[0032] (1) The electronic device can determine the interest characteristics of the preset user based on the user characteristics of the preset user and the item characteristics of each item among multiple items. Based on the user characteristics of the preset user and the user characteristics of each associated user among multiple associated users, the electronic device can determine the difference characteristics of the preset user. Based on the interest characteristics and difference characteristics of the preset user, the electronic device can train the initial item recommendation model to generate an item recommendation model with high prediction accuracy (i.e., the target item recommendation model). The electronic device can also train the initial user recommendation model based on the user characteristics of the preset user and the user characteristics of each user among multiple users to generate a user recommendation model with high prediction accuracy (i.e., the target user recommendation model). In this way, a fusion recommendation model with high prediction accuracy (i.e., the target fusion recommendation model) can be obtained. The electronic device can recommend items and users to users based on the target fusion recommendation model, which can improve the effectiveness of item recommendation and user recommendation.

[0033] (2) Since the first predicted result for the preset user is the item that the preset user is highly interested in, as predicted by the initial item recommendation model, and the first true result for the preset user is the actual item that the preset user is interested in, the first loss determined by the electronic device based on the first predicted result and the first true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Furthermore, since the second predicted result for the preset user is other users that the preset user is highly associated with, as predicted by the initial user recommendation model, and the second true result for the preset user is the actual associated user, the second loss determined by the electronic device based on the second predicted result and the second true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Then, based on the first loss and the second loss, the electronic device can obtain the target loss. Then, based on the target loss, the electronic device updates the parameters in the initial fusion recommendation model, which can conveniently and quickly generate the target fusion recommendation model and improve the efficiency of model training.

[0034] (3) The electronic device can perform feature recognition based on the user information of the preset user, the item identifier of each item and the user information of each associated user, so as to accurately and effectively obtain the user characteristics of the preset user, the item characteristics of each item and the user characteristics of each associated user.

[0035] (4) The electronic device can determine the potential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the preset user, the potential characteristics of the preset user can be determined. Finally, based on the potential characteristics of the preset user, the difference characteristics of the preset user can be accurately and effectively determined, and thus the model can be trained accurately.

[0036] (5) This application can determine the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users. Based on the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users, the interest characteristics and the difference characteristics of the target user can be obtained. The electronic device inputs the interest characteristics and the difference characteristics of the target user into the target item recommendation model, thereby accurately and effectively obtaining the recommended items for the target user and recommending the target object to the target user.

[0037] (6) Electronic devices can perform feature recognition based on the user information of the target user, the item identifier of each item and the user information of each associated user, so as to accurately and effectively obtain the user characteristics of the target user, the item characteristics of each item and the user characteristics of each associated user.

[0038] (7) The electronic device can determine the potential characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the target user, the potential characteristics of the target user can be determined. Finally, based on the potential characteristics of the target user, the associated characteristics of the target user can be accurately and effectively determined, thereby accurately training the model.

[0039] (8) The electronic device first determines the user characteristics of the target user and the user characteristics of each of the at least one users. The electronic device inputs the user characteristics of the target user and the user characteristics of each of the at least one users into the target user recommendation model, so as to accurately and effectively determine whether to recommend the other user to the target user. In the case that the second prediction result indicates that the target user is recommended to the other user, it means that the other user is the user that the target user is interested in. At this time, the electronic device can accurately and effectively recommend the other user to the target user.

[0040] It should be noted that the technical effects of any of the implementation methods in aspects four through eight can be found in the technical effects of the corresponding implementation methods in aspects one, two, and three, and will not be repeated here.

[0041] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0042] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application, and do not constitute an undue limitation of this application.

[0043] Figure 1 This is a flowchart illustrating a training method for a fusion recommendation model according to an exemplary embodiment;

[0044] Figure 2 This is a flowchart illustrating yet another training method for a fusion recommendation model according to an exemplary embodiment;

[0045] Figure 3 This is a flowchart illustrating yet another training method for a fusion recommendation model according to an exemplary embodiment;

[0046] Figure 4 This is a schematic diagram illustrating yet another training method for a fusion recommendation model according to an exemplary embodiment;

[0047] Figure 5 This is a schematic diagram illustrating the specific process of training a target item recommendation model according to an exemplary embodiment;

[0048] Figure 6 This is a flowchart illustrating an item recommendation method according to an exemplary embodiment;

[0049] Figure 7 This is a flowchart illustrating yet another item recommendation method according to an exemplary embodiment;

[0050] Figure 8 This is a flowchart illustrating yet another item recommendation method according to an exemplary embodiment;

[0051] Figure 9 This is a schematic diagram illustrating a specific process of recommending a target object to a target user according to an exemplary embodiment.

[0052] Figure 10 This is a schematic diagram illustrating the relationship between users and associated users according to an exemplary embodiment;

[0053] Figure 11 This is a schematic diagram illustrating the process of fusion recommendation and the training process of the fusion recommendation model according to an exemplary embodiment;

[0054] Figure 12This is a flowchart illustrating a user recommendation method according to an exemplary embodiment;

[0055] Figure 13 This is a block diagram illustrating a training apparatus for a fusion recommendation model according to an exemplary embodiment;

[0056] Figure 14 This is a block diagram illustrating an item recommendation device according to an exemplary embodiment;

[0057] Figure 15 This is a block diagram illustrating a user recommendation device according to an exemplary embodiment;

[0058] Figure 16 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Detailed Implementation

[0059] To enable those skilled in the art to better understand the technical solutions of this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0060] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0061] Furthermore, the terms “comprising” and “having”, and any variations thereof, used in the description of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include other steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0062] It should be noted that in the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0063] In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0064] As described in the background section, due to the rapid growth of internet data and the diversification of user behavior, methods such as statistical analysis and collaborative filtering may not be able to accurately identify items that users are interested in, thus affecting the effectiveness of item recommendations. Based on this, embodiments of this application provide a training method for a fusion recommendation model, an item recommendation method, and a user recommendation method. In these embodiments, an electronic device can determine the interest characteristics of a preset user based on the user characteristics of the preset user and the item characteristics of each item among multiple items. Furthermore, based on the user characteristics of the preset user and the user characteristics of each associated user among multiple associated users, the electronic device can determine the difference characteristics of the preset user. The electronic device trains an initial item recommendation model based on the interest characteristics and difference characteristics of the preset user, generating an item recommendation model (i.e., a target item recommendation model) with high prediction accuracy. The electronic device can also train an initial user recommendation model based on the user characteristics of the preset user and the user characteristics of each user among multiple users, generating a user recommendation model (i.e., a target user recommendation model) with high prediction accuracy. This leads to a fusion recommendation model (i.e., a target fusion recommendation model) with high prediction accuracy. The electronic device then uses this target fusion recommendation model to recommend items and users, thereby improving the effectiveness of item and user recommendations.

[0065] For example, the electronic device executing the training method and item recommendation method of the item recommendation model provided in this application embodiment can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery network (CDN) services, and big data and artificial intelligence platforms. This application does not impose special limitations on the specific form of the electronic device. It can interact with users through one or more methods such as a keyboard, touchpad, touch screen, remote control, voice interaction or handwriting device.

[0066] To facilitate understanding, the training method of the fusion recommendation model provided in this application will be described in detail below with reference to the accompanying drawings.

[0067] Figure 1 This is a flowchart illustrating a training method for a fusion recommendation model according to an exemplary embodiment, such as... Figure 1 As shown, the training method for this fusion recommendation model may include: S101-S104.

[0068] S101, The electronic device determines the user characteristics of a preset user, the item characteristics of each item among multiple items, and the user characteristics of each user among multiple users.

[0069] Among them, the multiple items are items that the preset user is interested in, and the multiple users include multiple associated users, which are users who have interactive behavior with the preset user.

[0070] In this embodiment of the application, the preset user can be one user or multiple users, and no limitation is made in this embodiment of the application.

[0071] Combination Figure 1 ,like Figure 2 As shown, in one implementation of this application embodiment, the above-mentioned determination of the user characteristics of the preset user, the item characteristics of each item among multiple items, and the user characteristics of each associated user among multiple associated users may specifically include S1011-S1012.

[0072] S1011, The electronic device obtains user information of a preset user, item identifier of each item, and user information of each associated user.

[0073] The user information of the preset user includes one or more of the following: the user's name, the user's age, and the user's location information.

[0074] It should be understood that the user information of each associated user is the same type of information or similar information as the user information of the preset user.

[0075] In one implementation of this application, the electronic device can determine the user profile of the preset user based on the preset user's name, age, gender, region, and hobbies, and the user profile can better determine the preset user's interests and preferences.

[0076] S1012 The electronic device performs feature recognition on the user information of the preset user, the item identifier of each item, and the user information of each associated user, respectively, to obtain the user characteristics of the preset user, the item characteristics of each item, and the user characteristics of each associated user.

[0077] In this embodiment of the application, the electronic device can perform feature recognition based on the user information of the preset user, the item identifier of each item, and the user information of each associated user, so as to accurately and effectively obtain the user characteristics of the preset user, the item characteristics of each item, and the user characteristics of each associated user.

[0078] S102. The electronic device determines the interest characteristics of the preset user based on the user characteristics of the preset user and the item characteristics of each item, and determines the difference characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each associated user among multiple associated users.

[0079] The preset user's interest features are used to characterize the preset user's level of interest in each item, and the preset user's difference features are used to characterize the degree of difference between the preset user and each associated user.

[0080] It should be understood that the electronic device can determine the interest characteristics of the preset user based on the user characteristics and item characteristics of the preset user. The interest characteristics are used to characterize the degree of interest of the preset user in each item. The electronic device can determine the difference characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each of the multiple associated users. The difference characteristics are used to characterize the degree of difference between the preset user and each of the associated users.

[0081] Combination Figure 1 ,like Figure 3 As shown, in one implementation of this application embodiment, the above-mentioned determination of the differential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each associated user may specifically include S1021-S1023.

[0082] S1021. The electronic device determines the potential characteristics of the preset user by multiplying the user characteristics of the preset user with the user characteristics of each associated user.

[0083] In one implementation of this application, the electronic device can multiply the user characteristics of the preset user with the user characteristics of each associated user. The electronic device can then obtain the result of the multiplication operation and determine the result as the potential characteristics of the preset user.

[0084] S1022. The electronic device determines the difference between the potential characteristics of the preset user and the user characteristics of the preset user as the associated characteristics of the preset user.

[0085] It should be understood that after the electronic device determines the potential characteristics of the preset user, it can determine the associated characteristics of the preset user based on the difference between the potential characteristics of the preset user and the user characteristics of the preset user.

[0086] S1023. The electronic device determines the square of the associated characteristics of the preset user as the differential characteristics of the preset user.

[0087] In this embodiment, the electronic device can determine the potential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the preset user, the potential characteristics of the preset user can be determined. Finally, based on the potential characteristics of the preset user, the difference characteristics of the preset user can be accurately and effectively determined, thereby accurately training the model.

[0088] In one implementation of this application, the electronic device can determine the interest characteristics of the preset user by multiplying the user characteristics of the preset user with the item characteristics of each item.

[0089] In another implementation of this application embodiment, the electronic device can determine that the differential characteristics of the preset user satisfy the following formula:

[0090]

[0091] in, T represents the differential feature vector of the preset user. u1,v1 U represents the product between user features and associated user features. v1 This represents the feature vector of the associated user.

[0092] S103. The electronic device inputs the preset user's interest features and the preset user's difference features into the initial item recommendation model to obtain the first prediction result of the preset user, and inputs the user features of the preset user and the user features of each user among multiple users into the initial user recommendation model to obtain the second prediction result of the preset user.

[0093] The initial item recommendation model is the item recommendation model included in the initial fusion recommendation model, and the initial user recommendation model is the user recommendation model included in the initial fusion recommendation model.

[0094] It should be understood that the first prediction result for the preset user is the item that the preset user is highly interested in, as predicted by the initial item recommendation model, and the second prediction result for the preset user is other users with a high degree of relevance, as predicted by the initial user recommendation model.

[0095] S104. The electronic device trains the initial fusion recommendation model based on the first prediction result of the preset user and the second prediction result of the preset user to generate the target fusion recommendation model.

[0096] The target fusion recommendation model includes a target item recommendation model and a target user recommendation model. The target item recommendation model is used to recommend items to a user, and the target user recommendation model is used to recommend other users to a user.

[0097] It should be understood that the initial item recommendation model is an untrained or initial state item recommendation model, and its prediction accuracy (or precision) may not be very high. The electronic device can train the initial item recommendation model to generate a target item recommendation model with higher accuracy. Similarly, the initial user recommendation model is an untrained or initial state user recommendation model, and its prediction accuracy (or precision) may not be very high. The electronic device can train the initial user recommendation model to generate a target user recommendation model with higher accuracy, thereby obtaining a target fusion recommendation model with higher accuracy.

[0098] In this embodiment, the electronic device can determine the interest characteristics of the preset user based on the user characteristics of the preset user and the item characteristics of each item among multiple items. Furthermore, based on the user characteristics of the preset user and the user characteristics of each associated user among multiple associated users, the electronic device can determine the difference characteristics of the preset user. Based on the interest characteristics and difference characteristics of the preset user, the electronic device trains an initial item recommendation model to generate an item recommendation model (i.e., a target item recommendation model) with high prediction accuracy. The electronic device can also train an initial user recommendation model based on the user characteristics of the preset user and the user characteristics of each user among multiple users to generate a user recommendation model (i.e., a target user recommendation model) with high prediction accuracy. This leads to a fusion recommendation model (i.e., a target fusion recommendation model) with high prediction accuracy. The electronic device then uses this target fusion recommendation model to recommend items and users, thereby improving the effectiveness of item and user recommendations.

[0099] Combination Figure 1 ,like Figure 4 As shown, in one implementation of this application embodiment, the initial fusion recommendation model is trained based on the first prediction result of the preset user and the second prediction result of the preset user to generate the target fusion recommendation model, which may specifically include S1041-S1044.

[0100] S1041, The electronic device obtains the first real result of the preset user and the second real result of the preset user.

[0101] It is understandable that the first true result for the preset user is the items that the preset user is interested in, and the second true result is the users associated with the preset user.

[0102] S1042, the electronic device determines a first loss based on the first true result of the preset user and the first prediction result of the preset user, and determines a second loss based on the second true result of the preset user and the second prediction result of the preset user.

[0103] The first loss is used to characterize the degree of inconsistency between the first true result of the preset user and the first predicted result of the preset user, and the second loss is used to characterize the degree of inconsistency between the second true result of the preset user and the second predicted result of the preset user.

[0104] It should be understood that the first prediction result of the preset user is the prediction result obtained after inputting the preset user's interest features and the preset user's difference features into the initial item recommendation model, and the second prediction result of the preset user is the prediction result obtained after inputting the preset user's user features and the user features of each of the multiple users into the initial user recommendation model.

[0105] S1043. The electronic device determines the target loss based on the first loss and the second loss.

[0106] It should be understood that the target loss is used to characterize the degree of inconsistency between the actual results of the pre-defined user fusion recommendation and the predicted results of the fusion recommendation.

[0107] S1044. The electronic device updates the parameters in the initial fusion recommendation model based on the target loss to obtain the target fusion recommendation model.

[0108] In one implementation of the application, the electronic device can iteratively update the parameters of the initial fusion recommendation model based on the target loss until the prediction accuracy (or precision) of the current fusion recommendation model is greater than or equal to the accuracy threshold. At this point, the electronic device can determine the current fusion recommendation model as the target fusion recommendation model.

[0109] In one implementation of this application, the electronic device can determine whether the fusion recommendation model after training is a high-precision model based on the evaluation results of the hit rate (HR) and normalized discounted cumulative gain (NDCG). If the fusion recommendation model is high-precision, it is determined to be the target fusion recommendation model.

[0110] In this embodiment, since the first predicted result for the preset user is the item with high interest predicted by the initial item recommendation model, and the first true result for the preset user is the actual item the preset user is interested in, the first loss determined by the electronic device based on the first predicted result and the first true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Furthermore, since the second predicted result for the preset user is other users with high relevance predicted by the initial user recommendation model, and the second true result for the preset user is the actual relevance of the preset user, the second loss determined by the electronic device based on the second predicted result and the second true result can accurately and effectively represent the difference between the result predicted by the initial item recommendation model and the true result. Then, based on the first loss and the second loss, the electronic device can obtain the target loss. Finally, based on the target loss, the electronic device updates the parameters in the initial fusion recommendation model, enabling convenient and quick generation of the target fusion recommendation model and improving the efficiency of model training.

[0111] The following example illustrates the specific process of training the target item recommendation model in this application.

[0112] For example, such as Figure 5 As shown, the electronic device first acquires multiple datasets and, based on these datasets, generates a user-item matrix and a user-related user matrix. Then, the electronic device can determine the user features of the user and the item features of the corresponding item. At this point, the electronic device can input the user-related user matrix into a convolutional network to generate the features of the related user. Based on the user's features, the item's features, and the related user's features, the user's interest features and differential features can be obtained. The model is trained based on the user's interest features and differential features to obtain the target item recommendation model.

[0113] Figure 6 This is a flowchart illustrating an item recommendation method according to an exemplary embodiment, such as... Figure 6 As shown, the recommended methods for this item may include: S201-S204.

[0114] S201, The electronic device determines the user characteristics of the target user, the item characteristics of each item in at least one item, and the user characteristics of each associated user in at least one associated user.

[0115] Wherein, the at least one item is an item that the target user may be interested in, and the at least one associated user is a user who has interacted with the target user.

[0116] Combination Figure 6 ,like Figure 7 As shown, in one implementation of this application embodiment, the above-mentioned determination of the user characteristics of the target user, the item characteristics of each item in at least one item, and the user characteristics of each associated user in at least one associated user may specifically include S2011-S2012.

[0117] S2011. The electronic device acquires the user information of the target user, the item identifier of each item, and the user information of each associated user.

[0118] The target user's information includes one or more of the following: the target user's username, the target user's age, and the target user's location information.

[0119] It should be understood that the user information of each associated user is the same type of information or similar information as the user information of the target user.

[0120] In one implementation of this application, the electronic device can determine the user profile of the target user based on the target user's name, age, gender, region, and hobbies, and the user profile of the target user can better determine the target user's interests and preferences.

[0121] S2012, The electronic device performs feature recognition on the user information of the target user, the item identifier of each item, and the user information of each associated user, respectively, to obtain the user characteristics of the target user, the item characteristics of each item, and the user characteristics of each associated user.

[0122] In this embodiment of the application, the electronic device can perform feature recognition based on the acquired user information of the target user, the item identifier of each item, and the user information of each associated user, thereby accurately and effectively obtaining the user characteristics of the target user, the item characteristics of each item, and the user characteristics of each associated user.

[0123] S202. The electronic device determines the target user's interest characteristics based on the target user's user characteristics and the item characteristics of each item, and determines the target user's difference characteristics based on the target user's user characteristics and the user characteristics of each associated user.

[0124] The target user's interest features are used to characterize the degree of interest the target user has in each item, and the target user's difference features are used to characterize the degree of difference between the target user and each associated user.

[0125] It should be understood that electronic devices can determine the target user's interest characteristics based on the target user's user characteristics and item characteristics. These interest characteristics are used to characterize the target user's degree of interest in each item. Electronic devices can also determine the target user's difference characteristics based on the target user's user characteristics and the user characteristics of each associated user. These difference characteristics are used to characterize the degree of difference between the target user and each associated user.

[0126] Combination Figure 6 ,like Figure 8 As shown, in one implementation of this application embodiment, the above-mentioned determination of the target user's differential characteristics based on the target user's user characteristics and the user characteristics of each associated user may specifically include: S2021-S2023.

[0127] S2021. The electronic device determines the potential characteristics of the target user by multiplying the user characteristics of the target user with the user characteristics of each associated user.

[0128] In one implementation of this application, the electronic device can multiply the user characteristics of the target user with the user characteristics of each associated user. The electronic device can then obtain the result of the multiplication operation and determine the result as the potential characteristics of the target user.

[0129] S2022. The electronic device determines the difference between the potential characteristics of the target user and the user characteristics of the target user as the associated characteristics of the target user.

[0130] It should be understood that after an electronic device identifies the potential characteristics of a target user, it can determine the associated characteristics of the target user based on the difference between the potential characteristics and the user characteristics of the target user.

[0131] S2023. The electronic device determines the differential characteristics of the target user by squared-up relevance features of the target user.

[0132] In this embodiment of the application, the electronic device can determine the potential characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user. Then, based on the potential characteristics and the user characteristics of the target user, the potential characteristics of the target user can be determined. Finally, based on the potential characteristics of the target user, the associated characteristics of the target user can be accurately and effectively determined, thereby accurately training the model.

[0133] In one implementation of this application, the electronic device can determine the target user's interest characteristics by multiplying the user characteristics of the target user with the item characteristics of each item.

[0134] In another implementation of this application embodiment, the electronic device can determine that the differential characteristics of the target user satisfy the following formula:

[0135]

[0136] in, T represents the differential feature vector of the target user. u2,v2 U represents the product of the target user's user characteristics and the associated user characteristics. v2 This represents the feature vector of the associated user.

[0137] S203. The electronic device inputs the target user's interest characteristics and the target user's difference characteristics into the target item recommendation model to obtain the target user's first prediction result.

[0138] The first prediction result for the target user includes the item identifier of the target item. The target item recommendation model is an item recommendation model included in the target fusion recommendation model, which is trained based on the training method of the fusion recommendation model provided in the embodiments of this application.

[0139] In this embodiment of the application, the target item recommendation model is a pre-trained item recommendation model with high accuracy, which is used to recommend items to a user.

[0140] S204. Electronic devices recommend target items to target users.

[0141] In one implementation of this application, the electronic device can send the item identifier of the target item to the target user's target terminal.

[0142] The following example illustrates the specific process by which an electronic device according to an embodiment of this application recommends a target object to a target user.

[0143] For example, such as Figure 9 As shown, the electronic device collects user data based on user information and saves the data to a cloud database. Based on the data in the cloud database, the electronic device can determine the user-item matrix and the user-related user matrix. Based on the data in the user-item matrix and the user-related user matrix and the target item recommendation model, the electronic device can obtain the target items that the user is highly interested in and recommend the target items to the target user.

[0144] The following example illustrates the association relationship of the user-associated user matrix in an embodiment of this application.

[0145] For example, such as Figure 10As shown, there are relationships between multiple users. Users directly associated with the target user are the target user's associated users. Arrows connecting users indicate one-way related users, and straight lines connecting users indicate two-way related users. The user feature similarity of two-way related users is greater than that of one-way related users.

[0146] In this embodiment, the electronic device first determines the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users. Based on the user characteristics of the target user, the item characteristics of each of the at least one item, and the user characteristics of each of the at least one associated users, the electronic device can obtain the target user's interest characteristics and the target user's difference characteristics. The electronic device inputs the target user's interest characteristics and the target user's difference characteristics into the target item recommendation model, thereby accurately and effectively obtaining the recommended items for the target user and recommending the target item to the target user.

[0147] The following example illustrates the specific process of item recommendation and the training of the item recommendation model in the embodiments of this application.

[0148] For example, such as Figure 11 As shown, the electronic device can extract samples from the database, extract features from these samples, and then generate a feature index based on the feature extraction results. The electronic device can then generate a training set from the extracted features and train a model based on this training set to obtain a fusion recommendation model. When a user uses the recommendation service, the electronic device first obtains the user's data based on service routing, extracts features according to the feature index, and inputs these features into the aforementioned fusion recommendation model to obtain recommended items or other users for the user.

[0149] Figure 12 This is a flowchart illustrating a user recommendation method according to an exemplary embodiment, such as... Figure 12 As shown, the recommended methods for this item may include: S301-S303.

[0150] S301, The electronic device determines the user characteristics of the target user and the user characteristics of each user among at least one user.

[0151] It is understandable that the at least one user is the one who determines whether to recommend the product to the target user.

[0152] S302. The electronic device inputs the user characteristics of the target user and the user characteristics of each user into the target user recommendation model to obtain the second prediction result of the target user.

[0153] The second prediction result user characterization of the target user indicates whether to recommend each user to the target user. The target user recommendation model is a user recommendation model included in the target fusion recommendation model. The target fusion recommendation model is trained based on the training method of the fusion recommendation model provided in the embodiments of this application.

[0154] In this embodiment of the application, the target user recommendation model is a user recommendation model that has been trained and has a high accuracy rate. This target user recommendation model is used to recommend items to a user.

[0155] S303. If the target user's second prediction result characterization recommends other users to the target user, the electronic device recommends other users to the target user.

[0156] Among them, the other users include at least one user included in the user group.

[0157] It should be understood that if the second prediction result for the target user indicates that other users are recommended to the target user, it means that the other users are users that the target user is interested in, and the electronic device can recommend the other users to the target user.

[0158] In this embodiment, the electronic device first determines the user characteristics of the target user and the user characteristics of each of the at least one users. The electronic device inputs the user characteristics of the target user and the user characteristics of each of the at least one users into the target user recommendation model, so as to accurately and effectively determine whether to recommend other users to the target user. If the second prediction result indicates that other users are recommended to the target user, it means that the other users are users that the target user is interested in. At this time, the electronic device can accurately and effectively recommend other users to the target user.

[0159] The foregoing mainly describes the solutions provided by the embodiments of this application from a methodological perspective. To achieve the above functions, the training device, item recommendation device, or electronic device for the item recommendation model includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0160] This application embodiment can, based on the above method, exemplarily divide the training device, item recommendation device, or electronic device of the item recommendation model into functional modules. For example, the training device, item recommendation device, or electronic device of the item recommendation model may include various functional modules corresponding to each functional division, or two or more functions may be integrated into one processing module. The integrated module can be implemented in hardware or as a software functional module. It should be noted that the module division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0161] Figure 13 This is a block diagram illustrating a training apparatus for a fusion recommendation model according to an exemplary embodiment. (Refer to...) Figure 13 The training device 100 for the fusion recommendation model includes: a determination unit 101, a processing unit 102, and a generation unit 103.

[0162] The determining unit 101 is used to determine the user characteristics of a preset user, the item characteristics of each item among multiple items, and the user characteristics of each user among multiple users. The multiple items are items that the preset user is interested in, and the multiple users include multiple associated users, which are users who have interactive behaviors with the preset user.

[0163] The determining unit 101 is used to determine the interest characteristics of the preset user based on the user characteristics of the preset user and the item characteristics of each item, and to determine the difference characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each of the multiple associated users. The interest characteristics of the preset user are used to characterize the degree of interest of the preset user in each item, and the difference characteristics of the preset user are used to characterize the degree of difference between the preset user and each associated user.

[0164] The processing unit 102 is used to input the interest features and difference features of the preset user into the initial item recommendation model to obtain the first prediction result of the preset user, and to input the user features of the preset user and the user features of each user into the initial user recommendation model to obtain the second prediction result of the preset user. The initial item recommendation model is an item recommendation model included in the initial fusion recommendation model, and the initial user recommendation model is a user recommendation model included in the initial fusion recommendation model.

[0165] The generation unit 103 is used to train the initial fusion recommendation model based on the first prediction result and the second prediction result of the preset user to generate a target fusion recommendation model. The target fusion recommendation model includes a target item recommendation model and a target user recommendation model. The target item recommendation model is used to recommend items to a user, and the target user recommendation model is used to recommend other users to a user.

[0166] Optionally, the training device 100 for the above-mentioned fusion recommendation model further includes: an acquisition unit 104.

[0167] The acquisition unit 104 is used to acquire the first real result and the second real result of the preset user.

[0168] The determining unit 101 is specifically used to determine a first loss based on the first true result of the preset user and the first prediction result of the preset user, and to determine a second loss based on the second true result of the preset user and the second prediction result of the preset user. The first loss is used to characterize the degree of inconsistency between the first true result of the preset user and the first prediction result of the preset user, and the second loss is used to characterize the degree of inconsistency between the second true result of the preset user and the second prediction result of the preset user.

[0169] The determining unit 101 is also specifically used to determine the target loss based on the first loss and the second loss.

[0170] The generation unit 103 is specifically used to update the parameters in the initial fusion recommendation model based on the target loss, so as to obtain the target fusion recommendation model.

[0171] Optionally, the training device 100 for the above-mentioned fusion recommendation model further includes: an acquisition unit 104.

[0172] The acquisition unit 104 is used to acquire user information of the preset user, item identifier of each item, and user information of each user. The user information of the preset user includes one or more of the following: user name of the preset user, user age of the preset user, and location information of the preset user.

[0173] The processing unit 102 is specifically used to perform feature recognition on the user information of the preset user, the item identifier of each item, and the user information of each user, respectively, so as to obtain the user features of the preset user, the item features of each item, and the user features of each user.

[0174] Optionally, the determining unit 102 is further specifically used to determine the product between the user characteristics of the preset user and the user characteristics of each associated user as the potential characteristics of the preset user;

[0175] The determining unit 102 is further specifically used to determine the difference between the potential characteristics of the preset user and the user characteristics of the preset user as the associated characteristics of the preset user;

[0176] The determining unit 102 is also specifically used to determine the square of the associated features of the preset user as the differential features of the preset user.

[0177] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0178] Figure 14 This is a block diagram illustrating an item recommendation device according to an exemplary embodiment. (Refer to...) Figure 14 The item recommendation device 200 includes: a determination unit 201 and a processing unit 202.

[0179] The determining unit 201 is used to determine the user characteristics of the target user, the item characteristics of each item in at least one item, and the user characteristics of each associated user in at least one associated user, wherein the at least one item is an item that the target user is interested in, and the at least one associated user is a user who has interactive behavior with the target user.

[0180] The determining unit 201 is configured to determine the target user's interest characteristics based on the user characteristics of the target user and the item characteristics of each item, and to determine the target user's difference characteristics based on the user characteristics of the target user and the user characteristics of each associated user. The target user's interest characteristics are used to characterize the target user's degree of interest in each item, and the target user's difference characteristics are used to characterize the degree of difference between the target user and each associated user.

[0181] Processing unit 202 is used to input the interest features and difference features of the target user into the target item recommendation model to obtain the first prediction result of the target user, which includes the item identifier of the target item;

[0182] Processing unit 202 is used to recommend the target item to the target user.

[0183] Optionally, the above-mentioned item recommendation device 200 further includes: an acquisition unit 203.

[0184] The acquisition unit 203 is used to acquire the user information of the target user, the item identifier of each item, and the user information of each associated user. The user information of the target user includes one or more of the following: the user name of the target user, the user age of the target user, and the location information of the target user.

[0185] The processing unit 202 is specifically used to perform feature recognition on the user information of the target user, the item identifier of each item, and the user information of each associated user, so as to obtain the user features of the target user, the item features of each item, and the user features of each associated user.

[0186] Optionally, the determining unit 201 is specifically used to determine the potential features of the target user as the product of the user features of the target user and the user features of each associated user.

[0187] The determining unit 201 is further specifically used to determine the difference between the potential characteristics of the target user and the user characteristics of the target user as the associated characteristics of the target user.

[0188] The determining unit 201 is also specifically used to determine the square of the associated features of the target user as the differential features of the target user.

[0189] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0190] Figure 15 This is a block diagram illustrating an item recommendation device according to an exemplary embodiment. (Refer to...) Figure 15 The item recommendation device 300 includes: a determination unit 301 and a processing unit 302.

[0191] The determining unit 301 is used to determine the user characteristics of the target user and the user characteristics of each user among at least one user.

[0192] The processing unit 302 is used to input the user features of the target user and the user features of each user into the target user recommendation model to obtain the second prediction result of the target user. The second prediction result of the target user is used to characterize whether to recommend each user to the target user. The target user recommendation model is a user recommendation model included in the target fusion recommendation model.

[0193] The processing unit 302 is configured to recommend other users to the target user when the second prediction result characterizes the target user as recommending other users to the target user, wherein the other users are users included in the at least one user.

[0194] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0195] Figure 16 This is a block diagram illustrating an electronic device according to an exemplary embodiment. Figure 16 As shown, the electronic device 400 includes, but is not limited to, a processor 401 and a memory 402.

[0196] The memory 402 described above is used to store the executable instructions of the processor 401. It is understood that the processor 401 is configured to execute instructions to implement the methods in the above embodiments.

[0197] It should be noted that those skilled in the art will understand that Figure 16 The electronic device structure shown does not constitute a limitation on the electronic device; the electronic device may include, but is not limited to, other electronic devices. Figure 16 This may indicate more or fewer components, or combinations of certain components, or different component arrangements.

[0198] Processor 401 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in memory 402, and by calling data stored in memory 302, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Processor 301 may include one or more processing units. Optionally, processor 401 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into processor 401.

[0199] The memory 402 can be used to store software programs and various data. The memory 402 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required by at least one functional module (such as a determination unit, processing unit, etc.), etc. Furthermore, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0200] In actual implementation, Figure 13 The functions of the determining unit 101, processing unit 102, generating unit 103, and acquiring unit 104 are as follows: Figure 14 The functions of the determining unit 201, processing unit 202, and acquiring unit 203 in the process are as follows: Figure 15 Both the determining unit 301 and the processing unit 302 can be determined by... Figure 16 The processor 401 calls the computer program stored in the memory 402 to implement the process. The specific execution process can be found in the description of the method section in the previous embodiment, and will not be repeated here.

[0201] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, such as a read-only memory (ROM), random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device.

[0202] In an exemplary embodiment, this application also provides a computer program product including one or more instructions, which can be executed by a processor 301 of an electronic device to perform the methods described above.

[0203] It should be noted that when one or more instructions in the computer-readable storage medium or computer program product are executed by the processor of an electronic device, they implement the various processes of the above method embodiments and achieve the same technical effect as the above method. To avoid repetition, they will not be described again here.

[0204] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0205] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0206] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the classified units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0207] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0208] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, essentially, or the part that contributes to the prior art, or a complete or partial classification of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0209] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A training method for a fusion recommendation model, characterized in that, include: The user characteristics of a preset user, the item characteristics of each item among multiple items, and the user characteristics of each user among multiple users are determined. The user characteristics of the preset user are determined based on a user profile used to characterize the interests and preferences of the preset user. The multiple items are items that the preset user is interested in, and the multiple users include multiple associated users, which are users who have interactive behaviors with the preset user. The interest characteristics of the preset user are determined based on the user characteristics of the preset user and the item characteristics of each item, and the difference characteristics of the preset user are determined based on the user characteristics of the preset user and the user characteristics of each of the multiple associated users. The interest characteristics of the preset user are used to characterize the degree of interest of the preset user in each item, and the difference characteristics of the preset user are used to characterize the degree of difference between the preset user and each associated user. The interest features and difference features of the preset user are input into the initial item recommendation model to obtain the first prediction result of the preset user. The first prediction result is the items that the preset user is highly interested in, as predicted by the initial item recommendation model. The user features of the preset user and the user features of each user among the multiple users are input into the initial user recommendation model to obtain the second prediction result of the preset user. The initial item recommendation model is the item recommendation model included in the initial fusion recommendation model, and the initial user recommendation model is the user recommendation model included in the initial fusion recommendation model. Based on the first prediction result and the second prediction result of the preset user, the initial fusion recommendation model is trained to generate a target fusion recommendation model. The target fusion recommendation model includes a target item recommendation model and a target user recommendation model. The target item recommendation model is used to recommend items to a user, and the target user recommendation model is used to recommend other users to a user.

2. The training method for the fusion recommendation model according to claim 1, characterized in that, The step of training the initial fusion recommendation model based on the first prediction result and the second prediction result of the preset user to generate the target fusion recommendation model includes: Obtain the first real result and the second real result of the preset user; A first loss is determined based on the first true result of the preset user and the first prediction result of the preset user, and a second loss is determined based on the second true result of the preset user and the second prediction result of the preset user. The first loss is used to characterize the degree of inconsistency between the first true result of the preset user and the first prediction result of the preset user, and the second loss is used to characterize the degree of inconsistency between the second true result of the preset user and the second prediction result of the preset user. Based on the first loss and the second loss, determine the target loss; Based on the target loss, the parameters in the initial fusion recommendation model are updated to obtain the target fusion recommendation model.

3. The training method for the fusion recommendation model according to claim 1, characterized in that, The process of determining the user characteristics of a preset user, the item characteristics of each item among multiple items, and the user characteristics of each user among multiple users includes: Obtain user information of the preset user, item identifier of each item, and user information of each user. The user information of the preset user includes one or more of the following: user name, user age, and location information of the preset user. Feature recognition is performed on the user information of the preset user, the item identifier of each item, and the user information of each user to obtain the user features of the preset user, the item features of each item, and the user features of each user.

4. The training method for the fusion recommendation model according to claim 1, characterized in that, The step of determining the differential characteristics of the preset user based on the user characteristics of the preset user and the user characteristics of each of the plurality of associated users includes: The product of the user characteristics of the preset user and the user characteristics of each associated user is determined as the potential characteristics of the preset user; The difference between the potential characteristics of the preset user and the user characteristics of the preset user is determined as the associated characteristics of the preset user; The square of the associated features of the preset user is determined as the differential feature of the preset user.

5. A method for recommending items, characterized in that, include: The user characteristics of the target user, the item characteristics of each item in at least one item, and the user characteristics of each associated user in at least one associated user are determined. The user characteristics of the target user are determined based on a user profile used to characterize the interests and preferences of the target user. The at least one item is an item that the target user is interested in, and the at least one associated user is a user who has interactive behavior with the target user. The target user's interest characteristics are determined based on the user characteristics of the target user and the item characteristics of each item, and the target user's difference characteristics are determined based on the user characteristics of the target user and the user characteristics of each associated user. The target user's interest characteristics are used to characterize the target user's degree of interest in each item, and the target user's difference characteristics are used to characterize the degree of difference between the target user and each associated user. The target user's interest features and the target user's difference features are input into the target item recommendation model to obtain the target user's first prediction result. The first prediction result is the item with a high degree of interest in the preset user predicted by the initial item recommendation model. The first prediction result of the target user includes the item identifier of the target item. The target item recommendation model is an item recommendation model included in the target fusion recommendation model. The target fusion recommendation model is trained based on the training method of the fusion recommendation model according to any one of claims 1-4. Recommend the target item to the target user.

6. The item recommendation method according to claim 5, characterized in that, The determination of the target user's user characteristics, the item characteristics of each item in at least one item, and the user characteristics of each associated user in at least one associated user includes: Obtain the user information of the target user, the item identifier of each item, and the user information of each associated user. The user information of the target user includes one or more of the following: the user name of the target user, the user age of the target user, and the location information of the target user. Feature recognition is performed on the user information of the target user, the item identifier of each item, and the user information of each associated user to obtain the user characteristics of the target user, the item characteristics of each item, and the user characteristics of each associated user.

7. The item recommendation method according to claim 5, characterized in that, The step of determining the differential characteristics of the target user based on the user characteristics of the target user and the user characteristics of each associated user includes: The product of the user characteristics of the target user and the user characteristics of each associated user is determined as the potential characteristics of the target user; The difference between the potential features of the target user and the user features of the target user is determined as the associated feature of the target user; The square of the associated features of the target user is determined as the differential feature of the target user.

8. A user recommendation method, characterized in that, include: Determine the user characteristics of the target user and the user characteristics of each user in at least one user group; The user features of the target user and the user features of each user are input into the target user recommendation model to obtain the second prediction result of the target user. The second prediction result of the target user is used to characterize whether to recommend each user to the target user. The target user recommendation model is a user recommendation model included in the target fusion recommendation model. The target fusion recommendation model is trained based on the training method of the fusion recommendation model according to any one of claims 1-4. If the second prediction result for the target user indicates that other users are recommended to the target user, then the other users are recommended to the target user, and the other users are users included in the at least one user.

9. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the training method of the fusion recommendation model as described in any one of claims 1 to 4, or the item recommendation method as described in any one of claims 5 to 7, or the user recommendation method as described in claim 8.

10. A computer-readable storage medium, characterized in that, When the computer-executable instructions stored in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is capable of executing the training method of the fusion recommendation model as described in any one of claims 1 to 4, or implementing the item recommendation method as described in any one of claims 5 to 7, or implementing the user recommendation method as described in claim 8.

Images