Object recommendation method, device, electronic device and storage medium

Abstract

The present disclosure relates to an object recommendation method, device, electronic device and storage medium, the method comprising: in response to an object recommendation request of a first account, obtaining operation log data of the first account; the operation log data represents the historical operation data of the historical operation object corresponding to the first account; obtaining multiple operation estimation data corresponding to multiple operation behaviors of each of multiple objects to be recommended; determining multiple weight information corresponding to the first account according to the operation log data; multiple weight information and multiple operation behaviors correspond one-to-one; based on multiple weight information and multiple operation estimation data corresponding to each object to be recommended, determining the ranking index data corresponding to each object to be recommended; based on the ranking index data corresponding to each object to be recommended, recommending objects to the first account. The present disclosure can realize personalized recommendations for different users, improve recommendation effects, and improve user experience and retention.

Description

Technical Field

[0001] The present disclosure relates to the field of Internet technology, and in particular to an object recommendation method, device, electronic device and storage medium. Background Art

[0002] The recommendation system is an information filtering system that filters a large amount of materials step by step through recall, sorting and other processes, and selects content that users are interested in for recommendation. In related technologies, when sorting candidate objects, they are usually sorted based on multi-objective estimation results given by the sorting model, such as effective play rate, long play rate, like rate, attention rate, etc. However, the sorting method in related technologies lacks personalization for the group, resulting in poor final recommendation results, which in turn affects user consumption experience and retention. Summary of the invention

[0003] The present disclosure provides an object recommendation method, device, electronic device and storage medium. The technical solution of the present disclosure is as follows:

[0004] According to a first aspect of an embodiment of the present disclosure, there is provided an object recommendation method, comprising:

[0005] In response to an object recommendation request of the first account, obtaining operation log data of the first account; the operation log data represents historical operation data of the historical operation object corresponding to the first account;

[0006] Obtaining a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended;

[0007] Determine multiple weight information corresponding to the first account according to the operation log data; the multiple weight information corresponds to the multiple operation behaviors one by one;

[0008] Determine the ranking index data corresponding to each object to be recommended based on the multiple weight information and the multiple operation estimation data corresponding to each object to be recommended;

[0009] Based on the ranking index data corresponding to each to-be-recommended object, an object is recommended to the first account.

[0010] In some possible embodiments, determining multiple weight information corresponding to the first account according to the operation log data includes:

[0011] Obtaining a vector space representation model;

[0012] The operation log data is input into a vector space representation model to obtain a feature vector corresponding to the first account, where the feature vector represents multiple weight information.

[0013] In some possible embodiments, the vector space representation model is generated by:

[0014] Acquire a training sample set; each training sample in the training sample set includes operation log data of the second account, a plurality of operation estimation data corresponding to the historical recommendation object recommended to the second account, and actual operation data performed by the second account on the historical recommendation object; the second account is any user account accessing the recommendation system;

[0015] Build pre-built machine learning models;

[0016] Inputting the operation log data of the second account into a preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes a plurality of prediction weight information; the plurality of prediction weight information corresponds one to one with the plurality of operation estimation data;

[0017] Determine the ranking index prediction data according to the prediction feature vector of the second account and the plurality of operation estimation data corresponding to the historical recommendation object;

[0018] Based on the ranking index prediction data and actual operation data, the preset machine learning model is trained to obtain the trained vector space representation model.

[0019] In some possible embodiments, the historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

[0020] In some possible embodiments, the format of each historical operation data is a binary format;

[0021] The identification information is represented by a binary number of a first preset number of digits, the operation behavior information is represented by a binary number of a second preset number of digits, and the object attribute information is represented by a binary number of a third preset number of digits.

[0022] In some possible embodiments, based on multiple weight information and multiple operation estimation data corresponding to each to-be-recommended object, determining the ranking index data corresponding to each to-be-recommended object includes:

[0023] Each operation estimation data in the plurality of operation estimation data is multiplied by the corresponding weight information and then summed up to obtain the ranking index data corresponding to each object to be recommended.

[0024] In some possible embodiments, recommending an object to the first account based on the ranking index data corresponding to each to-be-recommended object includes:

[0025] Sorting multiple objects to be recommended based on the ranking index data corresponding to each object to be recommended;

[0026] Determine a target recommendation object from the sorted multiple recommended objects;

[0027] The object recommendation feedback is sent to the terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

[0028] In some possible embodiments, obtaining a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended includes:

[0029] Get the estimated model;

[0030] Multiple operation behaviors are estimated for each object to be recommended according to the estimation model, so as to obtain multiple operation estimation data corresponding to each object to be recommended.

[0031] In some possible embodiments, the plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate;

[0032] The multiple weight information includes at least one of a like weight, a click weight, an effective play weight, a long play weight, and an attention weight corresponding to the operation estimation data.

[0033] According to a second aspect of an embodiment of the present disclosure, there is provided an object recommendation device, comprising:

[0034] A first acquisition module is configured to execute in response to an object recommendation request of the first account, and acquire operation log data of the first account; the operation log data represents historical operation data of the historical operation object corresponding to the first account;

[0035] A second acquisition module is configured to execute acquisition of a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended;

[0036] A first determination module is configured to determine a plurality of weight information corresponding to the first account according to the operation log data; the plurality of weight information corresponds to the plurality of operation behaviors one by one;

[0037] A second determination module is configured to determine the ranking index data corresponding to each to-be-recommended object based on the multiple weight information and the multiple operation estimation data corresponding to each to-be-recommended object;

[0038] The recommendation module is configured to perform object recommendation for the first account based on the ranking index data corresponding to each object to be recommended.

[0039] In some possible embodiments, the first determination module is further configured to execute acquisition of a vector space representation model; input the operation log data into the vector space representation model to obtain a feature vector corresponding to the first account, wherein the feature vector represents multiple weight information.

[0040] In some possible embodiments, a generation module of the vector space representation model is further included, and the generation module includes:

[0041] The acquisition submodule is configured to execute acquisition of a training sample set; each training sample in the training sample set includes operation log data of the second account, a plurality of operation estimation data corresponding to the historical recommendation object recommended to the second account, and actual operation data performed by the second account on the historical recommendation object; the second account is any user account accessing the recommendation system;

[0042] A construction submodule, configured to execute construction of a preset machine learning model;

[0043] The training submodule is configured to execute inputting the operation log data of the second account into a preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes multiple prediction weight information; the multiple prediction weight information corresponds one-to-one to multiple operation estimation data; the ranking index prediction data is determined according to the prediction feature vector of the second account and the multiple operation estimation data corresponding to the historical recommendation object; based on the ranking index prediction data and the actual operation data, the preset machine learning model is trained to obtain a trained vector space representation model.

[0044] In some possible embodiments, each of the plurality of historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

[0045] In some possible embodiments, the format of each historical operation data is a binary format;

[0046] The identification information is represented by a binary number of a first preset number of digits, the operation behavior information is represented by a binary number of a second preset number of digits, and the object attribute information is represented by a binary number of a third preset number of digits.

[0047] In some possible embodiments, the second determination module is further configured to perform multiplication of each operation estimation data in the plurality of operation estimation data with the corresponding weight information and then sum them up to obtain ranking index data corresponding to each object to be recommended.

[0048] In some possible embodiments, the sorting submodule is configured to sort the multiple objects to be recommended based on the sorting index data corresponding to each object to be recommended;

[0049] The sending submodule is configured to determine a target recommended object from the sorted multiple objects to be recommended, and send object recommendation feedback to the terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

[0050] In some possible embodiments, the second acquisition module is configured to execute acquisition of an estimation model; estimate multiple operation behaviors of each object to be recommended according to the estimation model to obtain multiple operation estimation data corresponding to each object to be recommended.

[0051] In some possible embodiments, the plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate;

[0052] The multiple weight information includes at least one of a like weight, a click weight, an effective play weight, a long play weight, and an attention weight corresponding to the operation estimation data.

[0053] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0054] processor;

[0055] a memory for storing processor-executable instructions;

[0056] The processor is configured to execute instructions to implement the object recommendation method provided in the first aspect of the embodiment of the present disclosure.

[0057] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the object recommendation method provided by the first aspect of the embodiment of the present disclosure.

[0058] According to the fifth aspect of an embodiment of the present disclosure, a computer program product is provided, the computer program product comprising a computer program, the computer program being stored in a readable storage medium, at least one processor of a computer device reading and executing the computer program from the readable storage medium, so that the computer device executes the object recommendation method provided in the first aspect of an embodiment of the present disclosure.

[0059] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0060] The object recommendation method of the disclosed embodiment, in the process of recommending to the first account, obtains the operation log data of the first account, and determines multiple weight information corresponding to the first account according to the operation log data, and the multiple weight information represents the preference of the first account for different operation behaviors; in this way, the differential needs of different users for different operation behaviors can be determined; then, for each object to be recommended, the multiple operation estimation data corresponding to the pre-estimated multiple operation behaviors are combined with the multiple weight information to obtain the ranking index data corresponding to each object to be recommended, and the ranking index data is used as the basis for the subsequent ranking of multiple objects to be recommended; the ranking index data integrates the user characteristics of the first account, so that the final ranking results of the multiple objects to be recommended are adapted to the personalized needs of different users, which can improve the user consumption experience and retention stickiness.

[0061] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0063] Figure 1 is a schematic diagram of an application environment according to an exemplary embodiment;

[0064] Figure 2 is a flowchart of an object recommendation method according to an exemplary embodiment;

[0065] Figure 3 is a flowchart showing a method of obtaining a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended according to an exemplary embodiment;

[0066] Figure 4 is a flow chart showing a method for generating a vector space representation model according to an exemplary embodiment;

[0067] Figure 5 is a flowchart showing a method of recommending an object to a first account according to an exemplary embodiment;

[0068] Figure 6 is a block diagram of an object recommendation device according to an exemplary embodiment;

[0069] Figure 7 It is a block diagram of an electronic device for object recommendation according to an exemplary embodiment. DETAILED DESCRIPTION

[0070] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0071] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar first objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0072] See also Figure 1 , which is a schematic diagram of an application environment of an object recommendation method according to an exemplary embodiment, the application environment may include a terminal 110 and a server 120, and the terminal 110 and the server 120 may be connected via a wired network or a wireless network.

[0073] The terminal 110 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The terminal 110 may be installed with client software such as an application (Application, referred to as App) that provides a human-computer interaction function. The application may be an independent application or a subprogram in the application. Exemplarily, the application may be a news application, a live broadcast application, or a video application. The user of the terminal 110 may log in to the application using pre-registered user information, which may include an account number and a password.

[0074] The server 120 may be a server that provides background services for the application in the terminal 110. Specifically, the service provided by the server 120 may be an object recommendation service, and the object may be determined according to the specific application scenario, for example, including but not limited to short videos, news information, advertisements, etc. The server 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0075] In a specific application scenario, the user information pre-registered by the user of the terminal 110 includes a first account, and the user can log in to the application through the first account to use the object recommendation service provided by the server 120; the terminal 110 can generate an object recommendation request based on a refresh operation triggered when the user refreshes the page, and send the object recommendation request to the server 120;

[0076] In response to the object recommendation request, the server 120 obtains the operation log data of the first account; the operation log data represents the historical operation data of the historical operation object corresponding to the first account in the historical period; wherein the historical operation object is the object operated by the first account, and the historical operation data includes the specific operation behavior performed by the first account on the object; the server 120 obtains multiple operation estimation data corresponding to multiple operation behaviors of each object to be recommended among multiple objects to be recommended; the object to be recommended is the object that the server 120 has preliminarily screened from a large number of object resources, and can also be called a recalled object; the server 120 determines multiple weight information corresponding to the first account according to the operation log data; multiple weight information corresponds to multiple operation behaviors one by one; then, the server 120 determines the ranking index data corresponding to each object to be recommended based on multiple weight information and multiple operation estimation data corresponding to each object to be recommended; finally, the server 120 recommends objects to the first account based on the ranking index data corresponding to each object to be recommended. The terminal 110 can display the objects recommended by the server 120 in the interface.

[0077] The object recommendation method of the disclosed embodiment, in the process of recommending to the first account, obtains the operation log data of the first account, and determines multiple weight information corresponding to the first account, and the multiple weight information represents the preference of the first account for different operation behaviors; in this way, the differential needs of different users for different operation behaviors can be determined; then, for each object to be recommended, the multiple operation estimation data corresponding to the pre-estimated multiple operation behaviors are combined with the multiple weight information to obtain the ranking index data corresponding to each object to be recommended, and the ranking index data is used as the basis for the subsequent ranking of multiple objects to be recommended; the ranking index data integrates the user characteristics of the first account, so that the final ranking results of the multiple objects to be recommended are adapted to the personalized needs of different users, which can improve the user consumption experience and retention stickiness.

[0078] It should be understood that Figure 1 The application environment shown is only an example. In actual applications, the object recommendation method of the embodiment of the present disclosure may be executed independently by a terminal or a server, or may be executed in cooperation with a terminal and a server. The embodiment of the present disclosure does not limit the specific application environment.

[0079] Figure 2is a flowchart of an object recommendation method according to an exemplary embodiment. Figure 2 As shown, the object recommendation method is used for Figure 1 Taking the server as an example, the following steps are included:

[0080] In step S201, in response to an object recommendation request of a first account, operation log data of the first account is obtained; the operation log data represents historical operation data of a historical operation object corresponding to the first account.

[0081] In the disclosed embodiment, the object recommendation request is used to instruct the server to recommend an object to the terminal that sent the request. After receiving the object recommendation request, the server completes the corresponding object recommendation service through the recommendation system. The object is the content that matches the actual application scenario; in some possible application scenarios, the object may include music, short videos, commodities, news information, advertisements, etc. The object recommendation request is generated at the terminal corresponding to the first account, and the first account may be any user account that has completed the registration operation on the server in advance; the terminal user logs in to the application through the pre-registered account, and the terminal generates an object recommendation request based on the user's preset operation on the application, and then sends the object recommendation request carrying its account information (first account) to the server. Thus, the server responds to the object recommendation request and obtains the operation log data of the first account; the operation log data represents the historical operation data of the historical operation object corresponding to the first account. The historical operation object refers to the object that has been operated by the first account, and the historical operation data is the data generated based on the operation of the first account on the historical operation object. Among them, the number of historical operation objects corresponding to the first account can be one or more.

[0082] When the number of historical operation objects corresponding to the first account is multiple, the historical operation objects corresponding to the first account may include all historical objects operated by the first account within the historical period, and accordingly, the historical operation data includes the specific operation behavior performed by the first account on each historical object; wherein the historical period can be set according to actual needs. For example, in a possible application scenario, the object is a short video, and the historical operation objects corresponding to the first account include all short videos operated by the first account in the past 3 days.

[0083] Alternatively, when the number of historical operation objects corresponding to the first account is multiple, the historical operation objects corresponding to the first account may include the most recent N objects operated by the first account, and accordingly, the historical operation data includes the specific operation behavior performed by the first account on each of the N objects. The specific value of N can be set according to actual needs.

[0084] In some possible embodiments, the historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

[0085] Specifically, the identification information represents the uniqueness of the historical operation object; the identification information may be in the form of a serial number, a character string, or a number, or a combination of multiple forms;

[0086] The operation behavior information includes the operation behavior actually performed by the first account on the historical operation object; the operation behavior information is related to the content of the object. For example, when the object is a short video, the operation behavior information may include at least one of playback information, like information, follow information, comment information, and click information of the publisher's personal homepage;

[0087] The object attribute information includes the attribute information of the historical operation object itself; for example, when the object is a short video, the object attribute information may include at least one of the length of the short video, the upload time, and the number of days from the upload time to today.

[0088] In some possible embodiments, the format of each historical operation data is a preset base format, such as binary format, octal format, decimal format or hexadecimal format; that is, each historical operation data is a long integer data, and when there are multiple historical operation objects, the multiple historical operation data corresponding to the multiple historical operation objects are represented in the form of a long integer data list; among them, taking the format of each historical operation data as binary format as an example, the number of binary bits of each historical operation data can be set according to actual needs.

[0089] Correspondingly, the identification information of the historical operation object can be represented by a binary number with a first preset number of digits, the operation behavior information can be represented by a binary number with a second preset number of digits, and the object attribute information can be represented by a binary number with a third preset number of digits.

[0090] In a specific embodiment, each historical operation data may include 64 binary bits, each binary bit has a value of 0 or 1, and the binary bit at the specified position is used to represent identification information, operation behavior information, and object attribute information; for example, the first binary bit of the 64 binary bits is designated to represent the like information in the operation behavior information, then when the bit is 0, it means that the first account has not liked the current short video, and when it is 1, it means that the first account has liked the current short video; for another example, the second binary bit of the 64 binary bits is designated to represent the attention in the operation behavior information, then when the bit is 0, it means that the first account has not followed the publisher of the current short video, and when it is 1, it means that the first account has followed the publisher of the current short video. In this way, the historical operation data of different historical operation objects are all represented in a unified form through the 64 binary bits.

[0091] In the above embodiment, the server responds to the object recommendation request of the first account, examines the operation log data of the first account, and represents the historical operation objects and corresponding historical operation data of the first account in a preset format to guide the preference for different operation behaviors in subsequent sorting, which can fill the gap in the utilization of user data by multiple operation estimation data in the sorting.

[0092] In step S203, a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended is obtained.

[0093] In the embodiment of the present disclosure, the server responds to the object recommendation request of the first account, and obtains multiple estimated operation data corresponding to multiple operation behaviors of each object to be recommended in multiple objects to be recommended; wherein the multiple objects to be recommended are a certain number of objects that the recommendation system determines from a massive object resource pool in the recall phase that the user may be interested in. The operation behavior can refer to the above embodiment, and the estimated operation data corresponding to the operation behavior characterizes the possibility of the user performing the operation behavior on the recommended object, that is, the specific definition of the estimated operation data is related to the content of the object. For example, when the object is a short video, the multiple estimated operation data corresponding to the multiple operation behaviors may include effective play rate (evtr), long play rate (lvtr), like rate (ltr), attention rate (wtr), etc.

[0094] In some possible embodiments, the acquisition of multiple estimated operation data corresponding to multiple operation behaviors of each of the multiple recommended objects in step S203 may include: Figure 3 The following steps are shown:

[0095] In step S301, an estimation model is obtained.

[0096] In step S303, multiple operation behaviors are estimated for each object to be recommended according to the estimation model to obtain multiple operation estimation data corresponding to each object to be recommended.

[0097] Among them, the estimation model is used to estimate multiple operation behaviors for each object to be recommended. The multiple operation estimation data corresponding to each object to be recommended output by the estimation model provide key information for the subsequent sorting of objects. The estimation accuracy of the estimation model for multiple operation estimation data will directly affect the final recommendation effect, and then affect the user consumption experience and retention.

[0098] In a specific embodiment, the estimation model is a single-target estimation model, that is, a model for estimating a single operation behavior. The single-target estimation model can be obtained by training through relevant technologies, and has a high estimation accuracy for a single target; therefore, in the above step S301, it specifically includes obtaining an estimation model for each operation behavior, such as obtaining a first operation behavior estimation model, a second operation behavior estimation model, a third operation behavior estimation model, etc.; correspondingly, in step 303, it specifically includes determining the first operation estimation data corresponding to the first operation behavior of each object to be recommended according to the first operation behavior estimation model, determining the second operation estimation data corresponding to the second operation behavior of each object to be recommended according to the second operation behavior estimation model, and determining the third operation estimation data corresponding to the third operation behavior of each object to be recommended according to the third operation behavior estimation model...; for example, in the short video application scenario, the above single-target estimation model includes a like rate estimation model and an attention rate estimation model. The click rate corresponding to each short video to be recommended is determined according to the like rate estimation model, and the attention rate corresponding to each short video to be recommended is determined according to the attention rate estimation model.

[0099] Further optionally, the estimated data output by the estimation model can be common to all users, that is, the estimated probability of each recommended object being performed a certain operation by different users is the same. The advantages of this model are simple structure and easy training, but the recommendation effect is not good in actual application; therefore, preferably, the estimated data output by the estimation model is personalized, that is, user features are integrated during training to obtain different possibilities for different users to perform a certain operation, which can further enhance the personalized recommendation effect.

[0100] In step S205, multiple weight information corresponding to the first account is determined according to the operation log data; the multiple weight information corresponds to the multiple operation behaviors one by one.

[0101] In step S207, based on the multiple weight information and the multiple operation estimation data corresponding to each object to be recommended, the ranking index data corresponding to each object to be recommended is determined.

[0102] In the disclosed embodiment, the server obtains multiple weight information corresponding to the first account from the operation log data of the first account, wherein the weight information represents the preference of the first account for a certain operation behavior; therefore, the multiple weight information corresponds one-to-one to the multiple operation behaviors; for each object to be recommended, the server fuses the corresponding multiple operation estimation data with the multiple weight information of the first account, and the fusion result is used as the ranking index data for each object to be recommended; the ranking index data is used as the basis for subsequent sorting of multiple objects to be recommended.

[0103] In actual business scenarios, the estimated business revenue is usually used as the basis for sorting multiple recommended objects. In order to ensure a higher business revenue as much as possible, the server sorts multiple recommended objects based on business revenue during the object recommendation process, so as to recommend the recommended objects with high business revenue to the user as the target recommended objects. The business revenue is determined based on multiple operation estimation data corresponding to multiple operation behaviors of the recommended object.

[0104] In the related technology, a weighted operation is performed on multiple operation estimate data corresponding to the recommended object, that is, different weights are assigned to each operation estimate data, and then the sum is calculated, and the result of the sum is used as the business benefit of the recommended object; for example, in the short video application scenario, the multiple operation estimate data of the short video to be recommended include effective play rate (evtr), long play rate (lvtr), like rate (ltr), and attention rate (wtr). The business benefit of the short video to be recommended can be represented as a*evtr+b*lvtr+c*ltr+d*wtr, wherein a represents the weight corresponding to the effective play volume, b represents the weight corresponding to the long play rate, c represents the weight corresponding to the like rate, and d represents the weight corresponding to the attention rate.

[0105] The problem in the related technology is that all users use the same weight list for weighting, that is, each user uses the above weights a, b, c and d to calculate the business revenue. This calculation method lacks user-side personalization and ignores the different preferences of different users for operation behaviors, which leads to inaccurate business revenue calculated in the end. For example, some users are more inclined to like, so a higher like weight can increase c*ltr, so that the recommended object can obtain a higher business revenue, and then can obtain a higher ranking in the subsequent sorting; however, some users have no knowledge of likes, and increasing the like weight will squeeze out the revenue brought by the remaining operation estimated data.

[0106] Therefore, in some possible embodiments, the above-mentioned determining the ranking index data corresponding to each to-be-recommended object based on the multiple weight information and the multiple operation estimation data corresponding to each to-be-recommended object may include the following steps:

[0107] Each operation estimation data in the plurality of operation estimation data is multiplied by the corresponding weight information and then summed up to obtain the ranking index data corresponding to each object to be recommended.

[0108] That is, the ranking index data corresponding to each object to be recommended is determined through a weighted sum fusion method.

[0109] In the short video application scenario, the plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate;

[0110] Correspondingly, the multiple weight information includes at least one of the like weight, click weight, effective play weight, long play weight, and attention weight corresponding to the operation estimation data.

[0111] Correspondingly, the ranking index data of a short video can be estimated like rate*like weight+estimated click rate*click weight+estimated effective play rate*effective play weight+estimated long play rate*long play weight+estimated attention rate*attention weight.

[0112] In the above embodiment, through the operation log data of each user, multiple weight information representing the different preferences for multiple operation behaviors can be obtained, so that in the subsequent steps, the different operation estimation data corresponding to the different operation behaviors are assigned corresponding weight information, so that the revenue corresponding to the user's preferred operation behavior accounts for a higher proportion in the final calculated business revenue, so that the business revenue guidance is used to achieve personalized sorting, which can enhance the user's loyalty, trust and positive experience for the brand or product, thereby improving user stickiness.

[0113] In some possible embodiments, determining the multiple weight information corresponding to the first account according to the operation log data may specifically include the following steps:

[0114] Obtaining a vector space representation model;

[0115] The operation log data is input into a vector space representation model to obtain a feature vector corresponding to the first account, where the feature vector represents multiple weight information.

[0116] Specifically, the preferences of different users for different operation behaviors are learned through a vector space representation model; the operation log data of the first account is passed through the vector space representation model to obtain multiple weight information corresponding to the different preferences of the first account for multiple operation behaviors. In this way, the differentiated needs of different users for different operation behaviors can be met.

[0117] In some possible embodiments, the generation method of the above-mentioned vector space representation model includes: Figure 4 The following steps are shown:

[0118] In step S401, a training sample set is obtained; each training sample in the training sample set includes operation log data of the second account, multiple operation estimation data corresponding to the historical recommendation objects recommended to the second account, and actual operation data performed by the second account on the historical recommendation objects.

[0119] The second account is any user account for accessing the recommendation system.

[0120] In step S403, a preset machine learning model is constructed.

[0121] In the model training phase, it is necessary to determine the structure of the model to be trained and collect a certain number of training samples for training the model to be trained.

[0122] In a specific embodiment, the model structure of the preset machine learning model can refer to the deep neural network (DNN) model structure in the relevant technology. The last layer of the model is the embedding layer, and the final output result is the embedding vector. The main purpose of the preset machine learning model is to learn the operation log data of different users to obtain the user's weight information for different operation behaviors; therefore, the dimension of the embedding vector finally output is the same as the number of operation behaviors, and the embedding vector represents the user's preference for each operation behavior.

[0123] To give a simple example, the operation log data of user A includes three historical operation data corresponding to three historical operation objects, each historical operation data includes two binary bits, the first binary bit represents the like information, and the second binary bit represents the follow information. According to the value-taking method in the above embodiment, the operation log data can be [00, 01, 10], where 00, 01, 10 respectively represent the three historical operation data; [00, 01, 10] is passed through the preset machine learning model, and the output result can be [a, b], where a represents the like weight, that is, the user A's preference for the like behavior, and b represents the follow weight, that is, the user A's preference for the follow behavior.

[0124] In step S405, the operation log data of the second account is input into a preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes multiple prediction weight information; the multiple prediction weight information corresponds one-to-one to the multiple operation estimation data.

[0125] In step S407, ranking index prediction data is determined based on the prediction feature vector of the second account and a plurality of operation estimation data corresponding to the historical recommendation objects.

[0126] In step S409, the preset machine learning model is trained based on the ranking index prediction data and the actual operation data to obtain a trained vector space representation model.

[0127] In the above steps S405 to S409, the preset machine learning model is trained according to the operation log data of the second account, the multiple operation estimation data and the actual operation data corresponding to the historical recommendation objects to obtain the trained vector space representation model.

[0128] The training sample includes a sample input and a label corresponding to the sample input. The sample input includes two parts, namely, the operation log data of the second account and a plurality of operation estimation data corresponding to the historical recommended objects recommended to the second account; wherein, the operation log data of the second account is used as the input of the preset machine learning model, and the output of the model represents the feature vector of the second account; here, the content of the operation log data of the second account can refer to the embodiment of the operation log data of the first account above, and will not be repeated here.

[0129] Another part of the sample input is multiple operation estimation data corresponding to the historical recommended objects recommended to the second account. This part is not used as the input of the preset machine learning model, but is fused with the output (feature vector) of the model. The fusion result is the ranking index prediction data (predicted business revenue) of the historical recommended objects; wherein, the historical recommended objects refer to the objects that the recommendation system has previously recommended to the second account, and the multiple operation estimation data corresponding to the historical recommended objects are determined by the estimation model during the previous recommendation process; further, the number of historical recommended objects can be one or more, and the multiple historical recommended objects can be all objects recommended by the recommendation system to the second account within the historical period, or the multiple historical recommended objects can also be the most recent N objects recommended by the recommendation system to the second account; wherein, the specific value of N can be set according to actual needs.

[0130] Correspondingly, the label corresponding to the sample input is the actual operation data performed by the second account on the historical recommendation object. The actual operation data is used to calculate the loss value with the above-mentioned fusion result (the ranking index prediction data of the historical recommendation object), and the parameters of the preset machine learning model are updated according to the loss value. After multiple iterations, a trained vector space representation model can be obtained.

[0131] In a specific embodiment, in the process of training a preset machine learning model to obtain a trained vector space representation model, the preset machine learning model can be determined as the current machine learning model; the operation log data of the second account is input into the current machine learning model to obtain a predicted feature vector corresponding to the second account; the ranking index prediction data is determined based on the predicted feature vector of the second account and multiple operation estimation data corresponding to the historical recommendation object; the loss value is determined based on the ranking index prediction data and the actual operation data; when the loss value is greater than a preset threshold, back propagation is performed based on the loss value, the current machine learning model is updated to obtain an updated machine learning model, and the updated machine learning model is re-determined as the current machine learning model; repeat the steps: input the operation log data of the second account into the current machine learning model to obtain a predicted feature vector corresponding to the second account; when the loss value is less than or equal to the preset threshold, the current machine learning model is determined as the vector space representation model.

[0132] Through the model training method introduced in the above embodiment, a vector space representation model can be generated, so that in actual applications, the vector space representation model can be used to convert the operation log data of different first accounts into personalized weight information for different operation behaviors; the personalized weight information is used to calculate the ranking index data of each object to be recommended, and the ranking index data is used as the basis for the subsequent ranking of multiple objects to be recommended. Therefore, different ranking results can be presented to different users, and personalized sorting can be achieved, which can improve the recommendation effect.

[0133] In addition, the improvement of recommendation effect helps users to continue to consume deeply, which in turn brings more and richer operation log data containing user behavior and user status, making the training of vector space representation model more accurate and forming a positive feedback loop.

[0134] In step S209, based on the ranking index data corresponding to each object to be recommended, an object is recommended to the first account.

[0135] In the disclosed embodiment, the server sorts multiple objects to be recommended according to the sorting index data corresponding to each object to be recommended determined in the above steps, and recommends objects to the first account according to the sorting result.

[0136] In some possible embodiments, the above-mentioned object recommendation for the first account based on the ranking index data corresponding to each object to be recommended may specifically include the following: Figure 5 The following steps are shown:

[0137] In step S501, multiple objects to be recommended are sorted based on the sorting index data corresponding to each object to be recommended.

[0138] In a specific embodiment, based on the numerical value of the ranking index data corresponding to each object to be recommended, multiple objects to be recommended are ranked in order from large to small or from small to large.

[0139] In step S503, a target recommended object is determined from the sorted plurality of objects to be recommended.

[0140] In a specific embodiment, after sorting the sorting index data from large to small according to their numerical values, the top N recommended objects are used as target recommended objects; wherein N is set according to actual needs.

[0141] In step S505, object recommendation feedback is sent to the terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

[0142] In a specific embodiment, the recommendation system of the server sends object recommendation feedback to the terminal corresponding to the first account, and the object recommendation feedback includes identification information of the target recommended object. The terminal corresponding to the first account can obtain the target recommended object from the object resource pool of the server according to the identification information of the target recommended object for display in the terminal interface.

[0143] In summary, in the embodiments of the present disclosure, when making a recommendation to the first account, the operation log data of the first account is obtained, and multiple weight information corresponding to the first account is determined according to the operation log data, and the multiple weight information represents the preference of the first account for different operation behaviors; in this way, the different needs of different users for different operation behaviors can be determined; then, for each object to be recommended, the multiple operation estimation data corresponding to the pre-estimated multiple operation behaviors are combined with the multiple weight information to obtain the ranking index data corresponding to each object to be recommended, and the ranking index data is used as the basis for the subsequent ranking of multiple objects to be recommended; the ranking index data integrates the user characteristics of the first account, so that the final ranking results of the multiple objects to be recommended adapt to the personalized needs of different users, which can improve the user consumption experience and retention stickiness.

[0144] Figure 6 is a block diagram of an object recommendation device according to an exemplary embodiment. Figure 6 The object recommendation device 600 includes a first acquisition module 610, a second acquisition module 620, a first determination module 630, a second determination module 640 and a recommendation model 650, wherein:

[0145] The first acquisition module 610 is configured to execute in response to the object recommendation request of the first account, and acquire the operation log data of the first account; the operation log data represents the historical operation data of the historical operation object corresponding to the first account;

[0146] The second acquisition module 620 is configured to execute acquisition of a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of the plurality of objects to be recommended;

[0147] The first determination module 630 is configured to determine a plurality of weight information corresponding to the first account according to the operation log data; the plurality of weight information corresponds to the plurality of operation behaviors one by one;

[0148] The second determination module 640 is configured to determine the ranking index data corresponding to each to-be-recommended object based on the multiple weight information and the multiple operation estimation data corresponding to each to-be-recommended object;

[0149] The recommendation module 650 is configured to perform object recommendation for the first account based on the ranking index data corresponding to each object to be recommended.

[0150] In some possible embodiments, the first determination module 630 is further configured to execute acquisition of a vector space representation model; input the operation log data into the vector space representation model to obtain a feature vector corresponding to the first account, where the feature vector represents multiple weight information.

[0151] In some possible embodiments, a generation module of the vector space representation model is further included, and the generation module includes:

[0152] The acquisition submodule is configured to execute acquisition of a training sample set; each training sample in the training sample set includes operation log data of the second account, a plurality of operation estimation data corresponding to the historical recommendation object recommended to the second account, and actual operation data performed by the second account on the historical recommendation object; the second account is any user account accessing the recommendation system;

[0153] A construction submodule, configured to execute construction of a preset machine learning model;

[0154] The training submodule is configured to execute inputting the operation log data of the second account into a preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes multiple prediction weight information; the multiple prediction weight information corresponds one-to-one to multiple operation estimation data; the ranking index prediction data is determined according to the prediction feature vector of the second account and the multiple operation estimation data corresponding to the historical recommendation object; based on the ranking index prediction data and the actual operation data, the preset machine learning model is trained to obtain a trained vector space representation model.

[0155] In some possible embodiments, each of the plurality of historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

[0156] In some possible embodiments, the format of each historical operation data is a binary format;

[0157] The identification information is represented by a binary number of a first preset number of digits, the operation behavior information is represented by a binary number of a second preset number of digits, and the object attribute information is represented by a binary number of a third preset number of digits.

[0158] In some possible embodiments, the second determination module 640 is further configured to perform a summation of multiplying each operation estimation data in the plurality of operation estimation data by the corresponding weight information to obtain ranking index data corresponding to each object to be recommended.

[0159] In some possible embodiments, the sorting submodule is configured to sort the multiple objects to be recommended based on the sorting index data corresponding to each object to be recommended;

[0160] The sending submodule is configured to determine a target recommended object from the sorted multiple objects to be recommended, and send object recommendation feedback to the terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

[0161] In some possible embodiments, the second acquisition module 620 is configured to execute acquisition of an estimation model; estimate multiple operation behaviors of each object to be recommended according to the estimation model, and obtain multiple operation estimation data corresponding to each object to be recommended.

[0162] In some possible embodiments, the plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate;

[0163] The multiple weight information includes at least one of a like weight, a click weight, an effective play weight, a long play weight, and an attention weight corresponding to the operation estimation data.

[0164] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0165] In an exemplary embodiment, an electronic device is also provided, including a processor; a memory for storing processor executable instructions; wherein the processor is configured to implement any one of the object recommendation methods provided in the embodiments of the present disclosure when executing the instructions stored in the memory.

[0166] The electronic device may be a terminal, a server or a similar computing device. For example, the electronic device is a server. Figure 7 is a block diagram of an electronic device for object recommendation according to an exemplary embodiment. Figure 7As shown, the server 700 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 710 (the processor 710 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 730 for storing data, and one or more storage media 720 (such as one or more mass storage devices) for storing application programs 723 or data 722. Among them, the memory 730 and the storage medium 720 can be short-term storage or permanent storage. The program stored in the storage medium 720 may include one or more modules, each of which may include a series of instruction operations on the server. Furthermore, the central processing unit 710 can be configured to communicate with the storage medium 720 and execute a series of instruction operations in the storage medium 720 on the server 700. The server 700 may also include one or more power supplies 770, one or more wired or wireless network interfaces 750, one or more input and output interfaces 740, and / or one or more operating systems 721, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0167] The input / output interface 740 may be used to receive or send data via a network. The specific example of the network may include a wireless network provided by a communication provider of the server 700. In one example, the input / output interface 740 includes a network adapter (Network Interface Controller, NIC), which may be connected to other network devices via a base station so as to communicate with the Internet. In one example, the input / output interface 740 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0168] It can be understood by those skilled in the art that Figure 7 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 7 More or fewer components as shown, or with Figure 7 Different configurations are shown.

[0169] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory 730 including instructions, and the instructions can be executed by the processor 710 of the device 700 to complete the above method. Alternatively, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0170] In an exemplary embodiment, a computer program product is also provided. The computer program is stored in a readable storage medium. At least one processor of a computer device reads and executes the computer program from the readable storage medium, so that the computer device executes any one of the object recommendation methods provided in the embodiments of the present disclosure.

[0171] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0172] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An object recommendation method, characterized in that: include: In response to an object recommendation request of a first account, obtaining operation log data of the first account; The operation log data represents historical operation data of the historical operation object corresponding to the first account; Obtaining a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended; Obtaining a vector space representation model; The vector space representation model obtains a plurality of prediction weight information by inputting the operation log data of the second account into a preset machine learning model, and fuses the plurality of prediction weight information with a plurality of operation estimation data corresponding to the historical recommendation object to obtain the ranking index prediction data of the historical recommendation object, and the preset machine learning model is trained based on the ranking index prediction data of the historical recommendation object and the actual operation data of the historical recommendation object; The second account is any user account for accessing the recommendation system; Inputting the operation log into the vector space representation model to obtain a feature vector corresponding to the first account, wherein the feature vector represents a plurality of weight information; the plurality of weight information corresponds to the plurality of operation behaviors one by one; The plurality of weight information represent the preference of the first account for different operation behaviors; Determine the ranking index data corresponding to each object to be recommended based on the multiple weight information and the multiple operation estimation data corresponding to each object to be recommended; Based on the ranking index data corresponding to each of the objects to be recommended, an object is recommended to the first account.

2. The object recommendation method according to claim 1, characterized in that: The method for generating the vector space representation model includes: Acquire a training sample set; each training sample in the training sample set includes operation log data of the second account, a plurality of operation estimation data corresponding to the historical recommendation object recommended to the second account, and actual operation data performed by the second account on the historical recommendation object; the second account is any user account accessing the recommendation system; Build pre-built machine learning models; Inputting the operation log data of the second account into the preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes a plurality of prediction weight information; the plurality of prediction weight information corresponds one to one with the plurality of operation estimation data; Determine ranking index prediction data according to the prediction feature vector of the second account and the plurality of operation estimation data corresponding to the historical recommendation object; Based on the ranking index prediction data and the actual operation data, the preset machine learning model is trained to obtain the trained vector space representation model.

3. The object recommendation method according to claim 1, characterized in that: The historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

4. The object recommendation method according to claim 3, characterized in that: The format of each historical operation data is binary format; The identification information is represented by a binary number with a first preset number of digits, the operation behavior information is represented by a binary number with a second preset number of digits, and the object attribute information is represented by a binary number with a third preset number of digits.

5. The object recommendation method according to claim 1, characterized in that: The determining, based on the plurality of weight information and the plurality of estimated operation data corresponding to each to-be-recommended object, the ranking index data corresponding to each to-be-recommended object comprises: Each operation estimation data in the plurality of operation estimation data is multiplied by the corresponding weight information and then summed to obtain the ranking index data corresponding to each object to be recommended.

6. The object recommendation method according to claim 1, characterized in that: The recommending an object to the first account based on the ranking index data corresponding to each to-be-recommended object includes: Sorting the multiple objects to be recommended based on the ranking index data corresponding to each object to be recommended; Determine a target recommendation object from the sorted multiple recommended objects; Sending object recommendation feedback to a terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

7. The object recommendation method according to claim 1, characterized in that: The obtaining of a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended includes: Get the estimated model; Multiple operation behaviors are estimated for each object to be recommended according to the estimation model to obtain multiple operation estimation data corresponding to each object to be recommended.

8. The object recommendation method according to any one of claims 1 to 7, characterized in that: The plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate; The multiple weight information includes at least one of a like weight, a click weight, an effective play weight, a long play weight, and an attention weight corresponding to the operation estimation data.

9. An object recommendation device, characterized in that: include: A first acquisition module is configured to execute in response to an object recommendation request of a first account and acquire operation log data of the first account; The operation log data represents historical operation data of the historical operation object corresponding to the first account; A second acquisition module is configured to execute acquisition of a plurality of operation estimation data corresponding to a plurality of operation behaviors of each of a plurality of objects to be recommended; A first determining module is configured to execute acquisition of a vector space representation model; The vector space representation model obtains multiple prediction weight information by inputting the operation log data of the second account into a preset machine learning model, and fuses the multiple prediction weight information with multiple operation estimation data corresponding to the historical recommendation object to obtain the ranking index prediction data of the historical recommendation object, and the preset machine learning model is trained based on the ranking index prediction data of the historical recommendation object and the actual operation data of the historical recommendation object; the second account is any user account that accesses the recommendation system; Inputting the operation log into the vector space representation model to obtain a feature vector corresponding to the first account, wherein the feature vector represents a plurality of weight information; the plurality of weight information corresponds to the plurality of operation behaviors one by one; The plurality of weight information represent the preference of the first account for different operation behaviors; A second determination module is configured to determine the ranking index data corresponding to each to-be-recommended object based on the plurality of weight information and the plurality of operation estimation data corresponding to each to-be-recommended object; The recommendation module is configured to perform object recommendation for the first account based on the ranking index data corresponding to each object to be recommended.

10. The object recommendation device according to claim 9, characterized in that: Also included is a generation module of the vector space representation model, the generation module comprising: an acquisition submodule, configured to execute acquisition of a training sample set; each training sample in the training sample set includes operation log data of a second account, a plurality of estimated operation data corresponding to a historical recommendation object recommended to the second account, and actual operation data performed by the second account on the historical recommendation object; the second account is any user account accessing the recommendation system; A construction submodule, configured to execute construction of a preset machine learning model; The training submodule is configured to input the operation log data of the second account into the preset machine learning model to obtain a prediction feature vector corresponding to the second account; the prediction feature vector includes multiple prediction weight information; the multiple prediction weight information corresponds one-to-one to the multiple operation estimation data; the ranking index prediction data is determined according to the prediction feature vector of the second account and the multiple operation estimation data corresponding to the historical recommendation object; based on the ranking index prediction data and the actual operation data, the preset machine learning model is trained to obtain the trained vector space representation model.

11. The object recommendation device according to claim 9, characterized in that: Each of the plurality of historical operation data includes at least one of identification information, operation behavior information, and object attribute information of a corresponding historical operation object.

12. The object recommendation device according to claim 11, characterized in that: The format of each historical operation data is binary format; The identification information is represented by a binary number with a first preset number of digits, the operation behavior information is represented by a binary number with a second preset number of digits, and the object attribute information is represented by a binary number with a third preset number of digits.

13. The object recommendation device according to claim 9, characterized in that: The second determination module is further configured to perform a multiplication of each operation estimation data in the plurality of operation estimation data with the corresponding weight information and then sum them up to obtain the ranking index data corresponding to each object to be recommended.

14. The object recommendation device according to claim 9, characterized in that: The recommendation module includes: A sorting submodule is configured to sort the multiple objects to be recommended based on the sorting index data corresponding to each object to be recommended; The sending submodule is configured to determine a target recommended object from the sorted multiple objects to be recommended, and send object recommendation feedback to the terminal corresponding to the first account; the object recommendation feedback includes identification information of the target recommended object.

15. The object recommendation device according to claim 9, characterized in that: The second acquisition module is configured to execute and acquire an estimation model; estimate multiple operation behaviors of each object to be recommended according to the estimation model, and obtain multiple operation estimation data corresponding to each object to be recommended.

16. The object recommendation device according to any one of claims 9 to 15, characterized in that: The plurality of operation estimation data include at least one of an estimated like rate, an estimated click rate, an estimated effective play rate, an estimated long play rate, and an estimated attention rate; The multiple weight information includes at least one of a like weight, a click weight, an effective play weight, a long play weight, and an attention weight corresponding to the operation estimation data.

17. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the object recommendation method according to any one of claims 1 to 8.

18. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the object recommendation method as claimed in any one of claims 1 to 8.

19. A computer program product, characterized in that The computer program product comprises a computer program, which is stored in a readable storage medium. At least one processor of a computer device reads and executes the computer program from the readable storage medium, so that the computer device performs the object recommendation method according to any one of claims 1 to 8.

Images