Recommended method, device, electronic equipment, storage medium and product

Abstract

The application discloses a recommendation method and device, electronic equipment, storage medium and product, which can perform coupling calculation on at least two types of portrait sets to obtain a first list, determine a matching relationship between X merchants and Y anchors according to the first list, and generate a target recommendation list according to the matching relationship. According to the embodiment of the application, the recommendation capability can be improved.

Description

Technical Field

[0001] This application belongs to the field of data processing technology, and in particular relates to recommended methods, apparatus, electronic devices, storage media and products. Background Technology

[0002] With the development of e-commerce live streaming, selling products through live streamers has become a way for merchants to quickly bring products into the public eye and rapidly increase sales while gaining traffic.

[0003] When merchants choose livestreamers to promote their products, they can only select them based on the livestreaming data of the platform. However, due to differences in industry and products, merchants may not be able to choose suitable livestreamers based on the livestreaming data. Moreover, with numerous livestreamers and merchants on the platform, it is also difficult for livestreamers to select suitable products from among the many products offered by various merchants. Therefore, it is impossible to generate appropriate recommendation data based on the livestreaming data of the platform, resulting in weak recommendation capabilities. Summary of the Invention

[0004] This application provides a recommendation method, apparatus, electronic device, storage medium, and product that can improve recommendation capabilities.

[0005] On the one hand, embodiments of this application provide a recommended method, which includes:

[0006] Coupled calculations are performed on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of X merchants, and the first recommendation list corresponding to each merchant includes the identity information of the streamer whose matching value with the merchant is greater than a first threshold. Where the at least two types of profile sets include a streamer profile set, a product profile set, and a user profile set, the first list includes a second recommendation list corresponding to each of Y streamers, and the second recommendation list corresponding to each streamer includes product information corresponding to products in the streamer's recommended product set whose product recommendation value is greater than a second threshold. X and Y are positive integers.

[0007] Based on the first list, determine the matching relationships between X merchants and Y live streamers, wherein the matching relationship includes at least one of the following: each of the X merchants is matched with at least one live streamer; each of the Y live streamers is matched with at least one merchant.

[0008] Based on the matching relationships, a target recommendation list is generated. Where the matching relationship includes at least one streamer matched by each of the X merchants, the target recommendation list includes the streamer recommendation list of each of the X merchants, and the streamer recommendation list of each merchant includes the identity information of at least one streamer matched by the merchant. Where the matching relationship includes at least one merchant matched by each of the Y streamers, the target recommendation list includes the merchant recommendation list of each of the Y streamers, and the merchant recommendation list of each streamer includes the identity information and product information of at least one merchant matched by the streamer.

[0009] On the other hand, embodiments of this application provide a recommended apparatus, which includes:

[0010] The coupling calculation module is used to perform coupling calculations on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of X merchants, and the first recommendation list corresponding to each merchant includes the identity information of the streamer whose matching value with the merchant is greater than a first threshold. Where the at least two types of profile sets include a streamer profile set, a product profile set, and a user profile set, the first list includes a second recommendation list corresponding to each of Y streamers, and the second recommendation list corresponding to each streamer includes product information corresponding to products in the streamer's recommended product set whose product recommendation value is greater than a second threshold. X and Y are positive integers.

[0011] The determination module is used to determine the matching relationship between X merchants and Y live streamers based on the first list, wherein the matching relationship includes at least one of the following: each of the X merchants is matched with at least one live streamer; each of the Y live streamers is matched with at least one merchant;

[0012] The generation module is used to generate a target recommendation list based on the matching relationship. Where the matching relationship includes at least one streamer matched by each of the X merchants, the target recommendation list includes the streamer recommendation list of each of the X merchants, and the streamer recommendation list of each merchant includes the identity information of at least one streamer matched by the merchant. Where the matching relationship includes at least one merchant matched by each of the Y streamers, the target recommendation list includes the merchant recommendation list of each of the Y streamers, and the merchant recommendation list of each streamer includes the identity information and product information of at least one merchant matched by the streamer.

[0013] In another aspect, embodiments of this application provide an electronic device, the device including: a processor and a memory storing computer program instructions. When the processor executes the computer program instructions, it implements the recommended method described above.

[0014] In another aspect, embodiments of this application provide a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the aforementioned recommended method.

[0015] In another aspect, embodiments of this application provide a computer program product in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform the aforementioned recommended method.

[0016] The recommendation method, apparatus, electronic device, storage medium, and product of this application embodiment can perform coupled calculations on at least two types of profile sets to obtain a first list, determine the matching relationship between X merchants and Y live streamers based on the first list, and generate a target recommendation list based on the matching relationship. Therefore, this application can couple calculations on at least two types of profile sets among the streamer profile set, product profile set, and user profile set to obtain a first recommendation list corresponding to each of the X merchants, so as to recommend streamers with a matching value greater than a first threshold to each merchant, and obtain a second recommendation list corresponding to each of the Y streamers, so as to recommend products in the streamer's recommended product set with a product recommendation value greater than the second threshold to each streamer. Based on the first list, the matching relationship between the X merchants and the Y streamers is determined to determine at least one streamer matched by each of the X merchants, or at least one merchant matched by each of the Y streamers. According to the matching relationship, a target recommendation list is generated. Streamers are recommended to each merchant through the streamer recommendation lists of the X merchants included in the target recommendation list, and merchants and their products are recommended to each streamer through the merchant recommendation lists of the Y streamers included in the target recommendation list. It is evident that the target recommendation list enables streamers to choose suitable products from among many merchants' products for promotion, and merchants to choose suitable streamers from among many streamers, thus improving recommendation capabilities. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a recommended method provided in an embodiment of this application;

[0019] Figure 2 This is a flowchart illustrating a recommended method provided in another embodiment of this application;

[0020] Figure 3 This is a schematic diagram of a broadcaster's historical live sales provided in one embodiment of this application;

[0021] Figure 4 This is a schematic diagram of the historical live-stream sales of a broadcaster provided in another embodiment of this application;

[0022] Figure 5 This is a schematic diagram of the historical live-stream sales of a broadcaster provided in another embodiment of this application;

[0023] Figure 6 This is a schematic diagram of the historical live-stream sales of a broadcaster provided in another embodiment of this application;

[0024] Figure 7 This is a schematic diagram of the historical live-stream sales of a broadcaster provided in another embodiment of this application;

[0025] Figure 8 This is a schematic diagram of the structure of a recommended device provided in an embodiment of this application;

[0026] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0027] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0028] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0029] To address the problems of the prior art, embodiments of this application provide recommended methods, apparatuses, electronic devices, storage media, and products. The recommended methods provided in the embodiments of this application are first described below.

[0030] Figure 1A flowchart illustrating a recommended method provided in an embodiment of this application is shown, as follows: Figure 1 As shown, the recommended method provided in this application embodiment includes the following steps:

[0031] Step 101: Perform coupled calculations on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of the X merchants. Each merchant's first recommendation list includes the streamer's identity information if the matching value with the merchant is greater than a first threshold. Where the at least two types of profile sets include a streamer profile set, a product profile set, and a user profile set, the first list includes a second recommendation list corresponding to each of the Y streamers. Each streamer's second recommendation list includes product information corresponding to products in the streamer's recommended product set whose product recommendation value is greater than a second threshold. X and Y are positive integers.

[0032] In the embodiments of this application, data is obtained from the database of the live-streaming host product sales platform. The data is preprocessed, such as through data cleaning, data standardization, and data field extraction. The preprocessed data or the original data of the live-streaming platform is used to construct user profiles for each user, host profiles for each host, and product profiles for each product. Users can be consumers, forming user profile sets, host profile sets, and product profile sets. Through self-coupling calculations of user profile sets, host profile sets, product profile sets and user profile sets, product profile sets and host profile sets, and user profile sets and host profile sets, a first recommendation list is generated for each merchant on the platform to show each merchant the matching degree between all hosts on the current live-streaming platform and themselves, making it easier for merchants to find hosts with high matching degrees. A second recommendation list is generated for each host on the platform to show each host the matching degree between all products on the current live-streaming platform and themselves, making it easier for hosts to find products with high matching degrees.

[0033] Step 102: Based on the first list, determine the matching relationship between X merchants and Y live streamers, wherein the matching relationship includes at least one of the following: each of the X merchants is matched with at least one live streamer; each of the Y live streamers is matched with at least one merchant.

[0034] In some embodiments, a first recommendation list is used to determine the top-ranked livestreamers who match each merchant, and a second list is used to determine the top-ranked products that match each livestreamer.

[0035] Step 103: Generate a target recommendation list based on the matching relationships. Where the matching relationships include at least one streamer matched by each of the X merchants, the target recommendation list includes the streamer recommendation list of each of the X merchants, and the streamer recommendation list of each merchant includes the identity information of at least one streamer matched by the merchant. Where the matching relationships include at least one merchant matched by each of the Y streamers, the target recommendation list includes the merchant recommendation list of each of the Y streamers, and the merchant recommendation list of each streamer includes the identity information and product information of at least one merchant matched by the streamer.

[0036] In some embodiments, the target recommendation list determines the information of the top-ranked streamers who match each merchant, and the target recommendation list also determines the information of the top-ranked merchants who match each streamer or the product information sold by those merchants. The information in the target recommendation list can then be recommended to the corresponding streamers and merchants to achieve streamer and merchant recommendations.

[0037] In embodiments of this application, at least two types of profile sets can be coupled and calculated to obtain a first list. Based on the first list, the matching relationship between X merchants and Y livestreamers is determined, and a target recommendation list is generated based on the matching relationship. Therefore, this application can couple and calculate at least two types of profile sets from the livestreamer profile set, product profile set, and user profile set to obtain a first recommendation list corresponding to each of the X merchants, so as to recommend livestreamers with a matching value greater than a first threshold to each merchant, and obtain a second recommendation list corresponding to each of the Y livestreamers, so as to recommend products from the livestreamer recommended product set with a product recommendation value greater than a second threshold to each livestreamer. Based on the first list, the matching relationship between the X merchants and Y livestreamers is determined to determine at least one livestreamer matched by each of the X merchants, or at least one merchant matched by each of the Y livestreamers. Based on the matching relationship, a target recommendation list is generated. Livestreamers are recommended to each merchant through the livestreamer recommendation lists of each of the X merchants included in the target recommendation list, and merchants and their products are recommended to each livestreamer through the merchant recommendation lists of each of the Y livestreamers included in the target recommendation list. It is evident that the target recommendation list enables livestreamers to select suitable products from numerous merchants for promotion, and allows merchants to choose suitable livestreamers from among many, thus improving recommendation capabilities.

[0038] In other embodiments of this application, when the first list is a second recommendation list, a first list is obtained by coupling calculation of at least two image sets, including:

[0039] Coupled calculations are performed on each first object portrait in the portrait set corresponding to the first object to obtain the first set; wherein, when the first object is a user, the first set is the user recommended product set, which includes: Q users and the recommended products for Q users; when the first object is a broadcaster, the first set is the broadcaster recommended product set, which includes: P broadcasters and the recommended products for P broadcasters, where Q and P are positive integers.

[0040] By coupling the user profile set and the product profile set, the first similarity matrix between users and products is obtained;

[0041] Using the users and products in the first similarity matrix whose similarity is greater than the third threshold, a set of products that users are interested in is generated. The set of products that users are interested in includes M users and M products that users are interested in, where M is a positive integer.

[0042] Coupled calculations are performed on the user profile set and the streamer profile set to obtain the user's favorite streamer set, which includes N users and N streamers that users like, where N is a positive integer;

[0043] If there is an overlap between the products in the user's recommended product set and the products in the recommended product set of the streamers in the user's favorite streamer set, or if there is an overlap between the products in the user's interested product set and the products in the recommended product set of the streamers in the user's favorite streamer set, the product recommendation value of the products in the streamer recommended product set will be increased by a preset value.

[0044] If there is no overlap between the products in the user-recommended product set and the products in the streamer-recommended product set of the streamers in the user's favorite streamer set, add the products in the user-recommended product set and the product recommendation values ​​of the products to the streamer-recommended product set.

[0045] If there is no overlap between the products in the set of products that the user is interested in and the products in the set of products recommended by the anchors that the user likes, then add the products in the set of products that the user is interested in and their product recommendation values ​​to the set of products recommended by the anchors.

[0046] Retrieve target products from the product recommendation set recommended by the livestreamer whose product recommendation value is greater than the fourth threshold;

[0047] Generate a second recommendation list, which includes product information for the target product.

[0048] In one example, when the first object is a user, user profiles of multiple users on the live streaming platform are obtained, resulting in a user profile set. Coupling calculations are performed between these user profiles to obtain multiple similar user pairs. Each user pair includes two users whose user profiles show high similarity in user data, such as similar interests, similar age, or similar types of streamers they follow. These multiple similar user pairs constitute a similar user pair set. The recommended purchase product list for each user on the live streaming platform is obtained using the relevant data from the users in the similar user pair set, and a user-recommended purchase product set is obtained from each user's recommended purchase product list. Similarly, streamer profiles of multiple streamers on the live streaming platform are obtained, resulting in a streamer profile set. Coupling calculations are performed between these streamer profiles to obtain multiple similar streamer pairs. Each streamer pair includes two streamers whose streamer profiles show high similarity in streamer data, such as similar fan types, similar age, or similar types of products they promote. These multiple similar streamer pairs constitute a similar streamer pair set. By analyzing relevant data from similar streamers, a list of recommended products for each streamer on the live streaming platform is obtained, and a set of recommended products is derived from each streamer's recommended product list. Next, user profiles of all users and product profiles on the live streaming platform are coupled to determine the products each user might be interested in, resulting in a set of products the user is interested in. Finally, user profiles of all users and streamer profiles on the live streaming platform are coupled to determine the streamers each user likes, resulting in a set of streamers the user prefers. For example, the top 10 streamers with the highest intimacy level between the user and the streamer can be considered the user's preferred streamers. For example, in the set of products recommended by users, identify the products recommended by user A. Then, in the set of broadcasters liked by user A, identify all broadcasters liked by user A. Next, identify the products recommended by each of these broadcasters. Determine if there is any overlap between the products recommended by each broadcaster and those recommended by user A. If there is, increment the recommendation value of the corresponding overlapping product by 1. If not, add user A's recommended product to the list of products recommended by each broadcaster and increment its recommendation value by 1. Similarly, in the set of products that user A is interested in, identify the products that user A is interested in. Then, in the set of broadcasters liked by user A, identify all broadcasters liked by user A. Next, identify the products recommended by each of these broadcasters and determine if there is any overlap between the products recommended by each broadcaster and those recommended by user A. If there is, increment the recommendation value of the corresponding overlapping product by 1. If not, add user A's interested product to the list of products recommended by each broadcaster and increment its recommendation value by 1.Obtain the product recommendation value for each product in the set of products recommended by the anchor. For products whose product recommendation value exceeds the fourth threshold or products whose product recommendation value ranks first, generate a second recommendation list. The second recommendation list includes: product name, product type, product recommendation value, etc.

[0049] It is understood that, through the embodiments of this application, the products recommended to the livestreamer are: products that users are interested in and products recommended for purchase by the user, thereby increasing the likelihood that users will purchase the products when the livestreamer recommends them.

[0050] In other embodiments of this application, coupling calculations are performed on each first object image in the image set corresponding to the first object to obtain a first set, including:

[0051] Coupled calculations are performed on the object profiles corresponding to each first object to determine the similarity between each first object;

[0052] Generate at least one first object pair, wherein each first object pair includes two first objects, and the similarity between the two first objects is greater than a fifth threshold;

[0053] Merge the first product lists of each first object in the first object pair to obtain the product set of the first object pair. Wherein, when the first object is a user and the first product list is the user's historical purchase product list, the product set is the user's historical purchase product set. When the first object is a live streamer and the first product list is the live streamer's historical sales products, the product set is the live streamer's historical sales product set.

[0054] Calculate the difference between the first product list and the product set of each first object in the first object pair to obtain the recommended product list of each first object.

[0055] Generate a first set, which includes each first object and a list of recommended products corresponding to each first object. The list of recommended products includes recommended products.

[0056] In one example, when the first object is a user, user profiles of multiple users on a live streaming platform are obtained to form a user profile set. Coupling calculations are performed between these user profiles to obtain multiple similar user pairs. Each user pair includes two users whose user profile data similarity is greater than a fifth threshold. The historical purchase product lists of the two users in each similar user pair are obtained. These lists are then merged to obtain the combined historical purchase product list of the two users. The difference between each user's historical purchase product list and the combined historical purchase product list is calculated to obtain a recommended purchase product list for each user. For example, suppose there exists a similar user pair (User 1, User 2). If User 1's historical purchase list is (Product 1, Product 2, Product 3) and User 2's historical purchase list is (Product 1, Product 3, Product 4), then the combined historical purchase list of this similar user pair is (Product 1, Product 2, Product 3, Product 4). Calculating the difference between this combined historical purchase list and User 1's historical purchase list reveals that User 1's recommended purchase product is Product 4; that is, User 1 will generate a list of (User 1, Product 4). Similarly, calculating the difference between this combined historical purchase list and User 2's historical purchase list reveals that User 2's recommended purchase product is Product 2; that is, User 2 will generate a list of (User 2, Product 2). Merging all users' recommended purchase lists forms the user-recommended product set. The same logic applies when the first user is a broadcaster, ultimately resulting in the broadcaster's recommended product set.

[0057] Understandably, by finding another similar user and analyzing the relationship between their past purchase lists and the user's past purchase lists, the system determines which products to recommend to that user. Because of the similarity between users, that user might also purchase products previously bought by another similar user, thus helping to recommend those products to the livestreamer later. Similarly, if livestreamers are similar and one livestreamer achieves high sales of a particular product, another similar livestreamer is likely to achieve similar sales, which can then help recommend that livestreamer to the merchant.

[0058] In other embodiments of this application, coupling calculations are performed on the object portraits corresponding to each first object to determine the similarity between the first objects, including:

[0059] Based on the first object information in the object profile corresponding to each first object, the feature vector of each first object is obtained;

[0060] Generate a first column vector, which includes the feature vectors of each first object;

[0061] Determine a first matrix based on a first column vector and the transpose of the first column vector. The first matrix includes Z*Z elements, and the element in the i-th row and j-th column represents the similarity between the i-th first object and the j-th first object, where Z, i, and j are positive integers, i ≤ Z, and j ≤ Z.

[0062] In some embodiments, when the first object is a user, a user portrait can be established from four dimensions: the user's basic information, the user's interest information, the user's behavior information, and the user's intimacy distribution information. The user's basic information may include: user gender, age group, geographical location, mobile phone model, operator information, etc., and can be directly obtained after data preprocessing. The user's interest information includes: the type distribution of the user's historical purchased goods, the type distribution of the watched live broadcasts, the style of the watched live streamers, etc. The user's interest information is statistical information and needs to be obtained through statistical analysis and calculation from the original live platform data. The user's behavior information includes: the frequency of the user watching live broadcasts, the average duration of each live broadcast watched, the reward records, the purchase records, the click purchase link records, etc. The user's behavior information is statistical information and needs to be obtained through statistical analysis and calculation from the original live platform data. The user's intimacy distribution information includes: the number of interactions with each live streamer, the number of rewards for each live streamer, the number of likes for each live streamer, the average duration of participating in each live streamer's live broadcast, the average number of orders placed per game in each live streamer's live broadcast, the average number of link clicks per game in each live streamer's live broadcast, etc. The user's intimacy distribution information is statistical information and needs to be obtained through statistical analysis and calculation from the original live platform data.

[0063] In one example, when the first object is a user, the first object information is the data in the basic information, interest information, and intimacy distribution information in the user portrait of the user, and a feature vector of the user is constructed based on the first object information. For example: the feature vector Xi of user i i Each dimension is as follows: (gender X i1 , age group X i2 , type of watched live broadcast X i3 , style of the watched live streamer X i4 , intimacy with streamer 1 X i5 , intimacy with streamer 2 X i6 , intimacy with streamer 3 X i7 ,..., intimacy with streamer n X in ). Obtain the feature vectors of all users and merge them to construct a first column vector X. For example: m is the total number of users on the platform. Calculate the norm of the first column vector X and merge it to construct a norm vector A of the first column vector: m represents the total number of users on the platform. Calculate the inner product between user feature vectors to form an inner product matrix P: P = X·X′, where X′ is the transpose of X. Calculate the inner product between the modulus vectors A to form an inner product matrix B: B = A·A′, where A′ is the transpose of A. Divide the corresponding elements of matrix P and matrix B to obtain the similarity matrix S between users. Each item S in the similarity matrix S ij This represents the similarity between user i and user j. A set of similar user pairs is constructed. Elements with a similarity greater than a specified threshold are extracted from the similarity matrix S. Based on the corresponding indices ij of each element, user i and user j are found, forming a pair (user i, user j). All found pairs constitute the set of similar user pairs.

[0064] In other embodiments, when the first target is a live streamer, a streamer profile can be established from four dimensions: the streamer's basic information, the streamer's live streaming information, the streamer's attractiveness, and the streamer's monetization ability. The streamer's basic information may include: gender, age, professional field, and education level, which can be directly obtained after data preprocessing. The streamer's live streaming information includes: live streaming style, areas of expertise, number of live streams, average live streaming duration, sales distribution by product type, and fan base. The live streaming style, areas of expertise, and number of live streams can be directly obtained after data preprocessing. The sales distribution by product type and the fan base are statistical information. The sales distribution by product type is obtained statistically from the streamer's sales data details; the fan base is obtained through statistical analysis of the streamer's fan data, and fan base information includes fan gender distribution, age distribution, geographical location distribution, mobile phone model distribution, and carrier distribution. The streamer's charm value information includes: number of streamer's fans, average number of users participating in each live stream, average user participation time, average number of likes received per live stream, average number of tips received per live stream, and average number of user interactions per live stream. The streamer's charm value information is statistical information and needs to be obtained from the original live streaming platform data through statistical analysis and calculation.

[0065] In one example, when the first object is a user, the information of the first object consists of the basic information of the broadcaster and the data in the broadcast information. Based on the information of the first object, a feature vector of the user is constructed, for example: feature vector X of broadcaster i. i The dimensions are as follows: (Gender X) i1 Age group X i2 Professional field X i3 Educational level X i4 Livestream Style X i5 Area of ​​expertise X i6 The percentage of male fans (X) i7 The percentage of fans by age group X i8 The percentage of fans by age group Xi9 The percentage of fans by age group X i10 The percentage of fans by age group X i11 The percentage of fans in different age groups is 5%. i12 The percentage of fans in different age groups is 6%. i13 The percentage of fans in different age groups is 7%. i14 Geographical distribution X ig Mobile phone model distribution X ip Carrier distribution X io ), where X ig X ip and X io This is a one-dimensional vector, representing the proportion of fans distributed by geographical location, mobile phone model, and carrier. Obtain the feature vectors of all streamers and merge them to construct the first column vector X. For example: m represents the total number of streamers on the platform. The modulus of the first column vector X is calculated, and the modulus vector A of the first column vectors is constructed by merging the results. m represents the total number of streamers on the platform. Calculate the inner product between the streamer feature vectors to form an inner product matrix P: P = X·X′, where X′ is the transpose of X. Calculate the inner product between the modulus vectors A to form an inner product matrix B: B = A·A′, where A′ is the transpose of A. Divide the corresponding elements of matrix P and matrix B to obtain the similarity matrix S between the streamers. Each item S in the similarity matrix S ij This refers to the similarity between broadcaster i and broadcaster j. A set of similar broadcaster pairs is constructed. Elements with similarity greater than a specified threshold are obtained from the similarity matrix S. Based on the corresponding index ij, broadcaster i and broadcaster j are found, forming a pair (broadcaster i, broadcaster j). All found pairs constitute the set of similar broadcaster pairs.

[0066] Understandably, by using user profiles and their self-coupling calculations, similar users can be found, making it easier to recommend products to them for purchase. Similarly, by using streamer profiles and their self-coupling calculations, similar streamers can be found, making it easier to recommend suitable streamers to merchants for purchase.

[0067] In other embodiments of this application, the user profile set and the streamer profile set are coupled and calculated to obtain a set of streamers favored by users, including:

[0068] Couple the user profile set and the anchor profile set to calculate the intimacy between each of the N users and the V anchors;

[0069] Generate a set of favorite broadcasters, which includes N users and V broadcasters among the N users whose intimacy with each user is greater than the sixth threshold.

[0070] In this embodiment, the purpose of coupling the streamer profile of each streamer in the live streaming platform with the user profile of each user in the live streaming platform is to associate users with streamers, generating a list of the top ten streamers with the highest intimacy level for each user. The elements in this set are a sequence of the form (User 1, Streamer 1, Streamer 2, Streamer 3, Streamer 4, Streamer 5, Streamer 6, Streamer 7, Streamer 8, Streamer 9, Streamer 10), where streamers 1 to 10 are arranged in descending order of user intimacy. The calculation process for coupling the streamer profile and the user profile is as follows: obtain the intimacy level in the user profile, which is based on intimacy distribution information; arrange the intimacy levels in descending order; obtain the top ten streamers with the highest intimacy levels, and establish the association between the user and these ten streamers, as well as the sequence between the user and the streamers.

[0071] Understandably, by coupling the user profile set and the streamer profile set, the streamers that each user likes can be determined, making it easier for merchants to find suitable streamers in the future.

[0072] In other embodiments of this application, when the first list is a first recommendation list, coupling calculations are performed on at least two image sets to obtain the first list, including:

[0073] By coupling the anchor profile set and the product profile set, a second similarity matrix between the anchor and the product is obtained;

[0074] Based on the products sold by the merchant, the merchant's feature vector is obtained;

[0075] Generate a second matrix, which includes the feature vectors of each merchant;

[0076] Based on the second similarity matrix and the second matrix, determine the matching degree matrix between the merchant and the live streamer;

[0077] Based on the matching degree matrix, target live streamers with a matching degree greater than the seventh threshold are identified;

[0078] Generate a first recommendation list, which includes the streamer information of the target streamer.

[0079] In this embodiment, the similarity matrix is ​​obtained by coupling the streamer profiles of all streamers and the product profiles of all products on the live streaming platform. The similarity matrix is ​​then used to determine the similarity between all streamers and all products. A merchant feature vector is generated for each merchant. Each dimension of the feature vector indicates whether the merchant sells a particular product; if so, that dimension is 1, otherwise 0. For example, if the live streaming platform contains product 1, product 2, and product 3, and merchants A, B, and C, and merchant A only sells product 1, then merchant A's vector would be (1, 0, 0). All merchant feature vectors constitute a second feature matrix M, where each element M... kj Let's determine whether merchant k sells product j. Calculate the inner product of the second feature matrix M and the similarity matrix R between the streamer and the product to obtain the matching degree matrix A between the merchant and the streamer: A = M·R′, where each item A in A... ki Let A be the match degree between merchant k and streamer j. Using the match degree matrix A, target streamers with a match degree greater than the seventh threshold are identified, and a first recommendation list is generated, which includes the streamer information of the target streamers.

[0080] Understandably, by generating a primary recommendation list to show each merchant the matching degree between themselves and all the streamers on the current live streaming platform, it is easier for merchants to find the streamers with the highest matching degree.

[0081] In another embodiment of this application, a first similarity matrix between users and products is obtained by coupling the user profile set and the product profile set; or, a second similarity matrix between broadcasters and products is obtained by coupling the broadcaster profile set and the product profile set, including:

[0082] Based on the information of the second object in the object profile corresponding to each second object, the feature vector of each second object is obtained, where the second object is either a user or a streamer;

[0083] Generate a second column vector, which includes the feature vectors of each second object;

[0084] Based on the product feature information in the product profile corresponding to each product, the feature vector of each product is obtained;

[0085] Generate a third column vector, which includes the feature vectors of each product;

[0086] Based on the second column vector, its transpose, the third column vector, and its transpose, the target similarity matrix is ​​determined. This target similarity matrix contains R*R elements. When the second object is a user, the target similarity matrix is ​​the first similarity matrix, where the element in row a and column b represents the similarity between the a-th user and the b-th product. When the second object is a broadcaster, the target similarity matrix is ​​the second similarity matrix, where the element in row a and column b represents the similarity between the a-th broadcaster and the b-th product. R, a, and b are positive integers, where a ≤ R and b ≤ R.

[0087] In some embodiments, data from a live streaming platform is acquired to create a product profile across three dimensions: basic product information, sales information, and product review distribution. Basic information includes: product name, product category, target audience, and product tags. This information is obtained directly after data preprocessing. Sales information includes: total sales volume, sales volume distribution by merchant, sales distribution by user age group, sales distribution by user gender, and sales distribution by region. This sales information is statistical and requires statistical analysis and calculation from the raw data. Product review distribution includes: product reviews distributed by merchant, product reviews distributed by user age group, and product reviews distributed by user gender. This product review distribution information is also statistical and requires statistical analysis and calculation from the raw data.

[0088] For example, when the second object is a user, the information of the second object consists of the user's basic information and interest information in the user profile. The user's feature vector is constructed using this second object information, and the feature vector of user i is X. i The dimensions are as follows: (Gender X) ix Age group X ia Occupation X ic Geographic location X ig Historical Purchased Product Types X is Among them, X ix X ia X ic X ig and X is This is a one-dimensional vector, representing the results of one-hot encoding for user gender, user age group, user occupation, user geographic location, and the distribution trend of historically purchased product types. The distribution trend of historically purchased product types is the normalized distribution law of historically purchased product types in the interest dimension of the user profile. The feature vectors of all users are merged to construct the second column vector X: m represents the total number of platform users. Calculate the modulus of the second column vector and combine them to construct the modulus vector A of the second column vector: m represents the total number of users on the platform. The basic information and sales information of the products in the product profile are used as product feature information, and a feature vector is constructed for each product. For example, the feature vector Y for product i... i The dimensions are as follows: (Gender bias in product sales Y) is The target age group for product sales tends to be Y. ia The occupation of product buyer Y ic Geographical distribution of product buyers Y ig Product type Y is ), where Y ix Y ia Y ic Y ig and Y is Let Y be a one-dimensional vector, representing the normalized distribution law of product sales by gender, the normalized distribution law of product sales by age group, the normalized distribution law of product buyers by occupation, the normalized distribution law of product buyers by geographical location, and the result of product type after one-hot encoding. The feature vectors of all products are merged to construct the third column vector Y. k represents the total number of products on the platform. Calculate the modulus of the third column vector and combine them to construct the modulus vector B of the third column vector: k represents the total number of products on the platform. Calculate the inner product between the user feature vector and the product feature vector, forming an inner product matrix P: P = X·Y′, where Y′ is the transpose of the third column vector Y. Calculate the inner product between the modulus vectors A and B, forming an inner product matrix L: L = A·B′, where B′ is the transpose of the modulus vector of the third column vector. Divide the corresponding elements of matrix P and matrix L to obtain the similarity matrix S between each user and each product. Each item S in the similarity matrix S ij This represents the similarity between user i and product j. A set of similar user-product pairs is constructed. Elements with similarity greater than a specified threshold are extracted from the similarity matrix S. Based on the corresponding index ij, user i and product j are found, forming a tuple (user i, product j), indicating that user i may be interested in product j. Finally, the set of products that the user is interested in is obtained.

[0089] In one example, when the second object is a live streamer, the information of the second object is the sales distribution by product type and fan group data from the live stream information in the live streamer profile. The feature vector of the live streamer is constructed using this second object information. The feature vectors of all live streamers are merged to construct the second column vector X. m represents the total number of streamers on the platform. Calculate the modulus of the second column vector and combine them to construct the modulus vector A of the second column vector: m represents the total number of livestreamers on the platform. The basic information and sales information of the products in the product profile are used as product feature information, and a feature vector is constructed for each product. For example, the feature vector Y for product i... i The dimensions are as follows: (Gender bias in product sales Y) is The target age group for product sales tends to be Y. ia The occupation of product buyer Y ic Geographical distribution of product buyers Y ig Product type Y is ), where Y ix Y ia Y ic Y ig and Y is Let Y be a one-dimensional vector, representing the normalized distribution law of product sales by gender, the normalized distribution law of product sales by age group, the normalized distribution law of product buyers by occupation, the normalized distribution law of product buyers by geographical location, and the result of product type after one-hot encoding. The feature vectors of all products are merged to construct the third column vector Y. k represents the total number of products on the platform. Calculate the modulus of the third column vector and combine them to construct the modulus vector B of the third column vector: k represents the total number of products on the platform. Calculate the inner product between the anchor feature vector and the product feature vector, forming the inner product matrix P: P = X·Y′, where Y′ is the transpose of the third column vector Y. Calculate the inner product between the modulus vector A and the modulus vector B, forming the inner product matrix L: L = A·

[0090] B′, where B′ is the transpose of the modulus vector of the third column vector. Dividing the corresponding elements of matrix P and matrix L yields the similarity matrix R between each broadcaster and each product. Each item R in the similarity matrix R ij This refers to the similarity between broadcaster i and product j, which is also defined here as the matching degree between broadcaster and product.

[0091] Understandably, in this embodiment, by determining the similarity between the streamer and the product, and the similarity between the user and the product, it is convenient to recommend merchants to the streamer and to recommend streamers to merchants in the future.

[0092] In other embodiments of this application, the streamer profile includes the streamer's monetization potential or streamer charm value, and the user profile includes the intimacy between the user and each streamer. Before coupling and calculating at least two types of profiles to obtain the first list, the method may further include:

[0093] The streamer's charm value is obtained by statistically analyzing and calculating the number of fans, the average number of users participating in each live stream, the average duration of user participation, the average number of likes received per live stream, the average number of tips received per live stream, and the average number of user interactions per live stream.

[0094] The monetization power of livestreamers is obtained by statistically analyzing and calculating total sales, average sales per session, average number of orders per session, average number of link clicks per session, average number of orders per user per session, and average sales per user per session.

[0095] The system statistically analyzes and calculates the number of times users interact with each streamer, the number of times they tip each streamer, the number of times they like each streamer, the average duration of their participation in each streamer's live stream, the average number of orders placed during each streamer's live stream, and the average number of links clicked during each streamer's live stream to determine the intimacy level between users and each streamer.

[0096] In this embodiment, the streamer's charm value is calculated from the streamer's charm value-related information. The calculation process is as follows: Weights are set for various dimensions affecting the charm value calculation, resulting in a weight vector W. The weight vector consists of: (number of fans weight W1, average number of users participating in each live stream weight W2, average user participation time weight W3, average number of likes received per stream weight W4, average number of rewards received per stream weight W5, average number of user interactions per stream weight W6). A statistical value vector S is calculated for all dimensions across the platform. The statistical value vector S consists of: (total number of fans on the platform S1, average number of users participating in each live stream on the platform S2, average user participation time on the platform S3, average number of likes received per stream on the platform S4, average number of rewards received per stream on the platform S5, average number of user interactions per stream on the platform S6). Calculate the statistical value vector H for various dimensions of the streamer's information. Each item in the statistical value vector H is: (Number of streamer's fans H1, Average number of users participating in each live stream H2, Average user participation time H3, Average number of likes received per live stream H4, Average number of tips received per live stream H5, Average number of user interactions per live stream H6). Calculate the streamer's charm value C using the following formula:

[0097] The monetization power of a streamer is calculated from relevant information about the streamer's monetization power. The calculation process is as follows: set the weights of various dimensions that affect monetization power in the calculation, and obtain the weight vector W. The weight vector consists of the following components: (total sales revenue weight W1, average sales revenue weight W2, average order quantity weight W3, average link clicks weight W4, average orders per user weight W5, and average sales revenue per user weight W6). Calculate the statistical value vector S for all dimensions of information across the platform. The components of statistical value vector S are: (total platform sales revenue S1, average platform sales revenue S2, average order quantity S3, average link clicks S4, average orders per user S5, and average sales revenue per user S6). Calculate the statistical value vector H for all dimensions of information about the streamer. The components of statistical value vector H are: (total streamer sales revenue H1, average sales revenue H2, average order quantity H3, average link clicks H4, average orders per user H5, and average sales revenue per user H6). Calculate the streamer's monetization power L using the following formula:

[0098] The intimacy level between a user and each streamer is represented by an intimacy vector, calculated based on the user intimacy distribution information in the user profile. The values ​​of each dimension in the vector represent the intimacy level between the user and each streamer. The intimacy calculation process involves setting weights for various dimensions affecting intimacy, resulting in a weight vector W. The weight vector consists of: (interaction frequency weight, tipping frequency weight, like frequency weight, duration weight, order count weight, link click count weight). Statistical value vectors S and R are calculated for each dimension. Taking user i and streamer j as an example, the statistical value vector S consists of: (number of interactions between user i and streamer j S1, number of tips from user i to streamer j S2, number of likes from user i to streamer j S3, average duration of user i's participation in streamer j's live stream S4, average number of orders from user i's participation in streamer j S5, average number of link clicks from user i's participation in streamer j S6).

[0099] The proportion vector R is obtained by normalizing the statistical value vector S:

[0100]

[0101] Where n is the total number of streamers on the platform. Calculate the intimacy vector I between users and each streamer: I = W·R′, where R′ is the transpose of the proportional vector R.

[0102] Understandably, in this embodiment, the intimacy between users and each streamer is calculated using the user intimacy distribution information in the user profile, the streamer's charm value is obtained from the streamer's charm value, and the streamer's monetization ability is obtained from the streamer's monetization ability. This can help merchants more intuitively judge the streamer's sales performance and help merchants rank streamers based on the intimacy between users and each streamer, charm value, and monetization ability, so as to find suitable streamers.

[0103] In other embodiments of this application, determining the matching relationship between X merchants and Y livestreamers based on a first list includes:

[0104] Given that the first list is the second recommendation list, obtain the distribution of merchant reviews in the product information of each Y live streamer in the second recommendation list;

[0105] Identify at least one merchant that matches each of the Y streamers, where at least one merchant is one whose merchant rating distribution is greater than the ninth threshold.

[0106] In this embodiment, after recommending products to the streamer through the second recommendation list, since there are many products on the live streaming platform, there may be multiple merchants selling this product. Therefore, the distribution of reviews of all merchants selling this product allows the streamer to find merchants with high user reviews after the product is sold and recommend these merchants to the streamer. This allows the streamer to find products suitable for their own sale, and the product quality can be guaranteed to a certain extent.

[0107] In other embodiments of this application, determining the matching relationship between X merchants and Y livestreamers based on a first list includes:

[0108] If the first list is the first recommendation list, obtain at least one of the following from the streamer information in the first recommendation list of each of the X merchants: streamer monetization ability or streamer charm value.

[0109] Identify at least one streamer who matches each of the X merchants, where at least one streamer is a streamer whose monetization ability or charm value is greater than the tenth threshold.

[0110] In this embodiment, the higher the streamer's charm value, the more fans they may have, and the greater the likelihood that fans will purchase the products promoted by that streamer. Therefore, the streamer's monetization potential is also higher. To better sell merchants' products and increase sales, streamers with high monetization potential or charm values ​​can be found in the first recommendation list and recommended to merchants to increase the sales volume of the products they sell.

[0111] In other embodiments of this application, after generating a target recommendation list based on the degree of matching, the method may further include:

[0112] Input the target recommendation list into the product sales revenue prediction model to obtain the predicted average sales volume of the products in the target recommendation list. The product sales revenue prediction model is trained using historical data of the products.

[0113] Calculate the difference between the predicted average sales volume and the average sales volume in historical data to obtain the product's evaluation value;

[0114] If the evaluation value is greater than or equal to the eleventh threshold, a formal recommendation list is generated.

[0115] In this embodiment, the product sales revenue prediction model is built using historical product data from the live streaming platform. The product sales revenue prediction model can be: y = ∑a i x i +b, y represents the streamer's sales volume, x i a represents the average conversion rate of product i. i 'b' represents the weighting coefficient, and 'b' represents the error. After the product sales revenue prediction model is established, the list of products sold in the previous live stream is obtained to verify the accuracy of the model. If the model is verified to be accurate or has a small error, the target recommendation list is input into the model to obtain the predicted average sales volume for each product. The difference between the predicted average sales volume of a product and the average sales volume of that product in the previous live streams is calculated; this difference is the evaluation value of the product's recommendation effect. If the evaluation value is greater than the eleventh threshold, the target recommendation list is converted into the official recommendation list.

[0116] It's understandable that by inputting each product in the target recommendation list into the product sales revenue prediction model, merchants can predict the sales performance of their products after using the livestreamer, thus helping them find suitable livestreamers. It can also help livestreamers predict the sales performance of their recommended products, thus helping them find suitable merchants and improving their recommendation capabilities.

[0117] Figure 2 A flowchart illustrating a recommended method provided in another embodiment of this application is shown, as follows: Figure 2 As shown, data is first collected from the live streaming platform, processed, and then user profile sets, streamer profile sets, and product profile sets are constructed. After construction, coupled calculations are performed on the profiles to obtain the target recommendation list. The target recommendation list is then input into the product sales revenue prediction model to obtain the official recommendation list, and the recommendation results are output to each merchant and streamer on the live streaming platform.

[0118] The following specific embodiment illustrates this application, assuming that the operator sells the following products:

[0119]

[0120]

[0121]

[0122]

[0123] Five live streamers were selected from the business hall and distributed as follows:

[0124] Serial Number Name districts and counties Department / Work Group / Channel 1 Zheng Jiao Eastern Marketing Center Important grid 2 Su Huijun Panji Marketing Center Government and Enterprise Customer Department 3 Jin Xin Panji Marketing Center Panji urban area grid 4 Li Dongdong Western Marketing Center Marketing Department 5 Wang Ting Western Marketing Center Government and Enterprise Customer Department

[0125] Figure 3-7 This application illustrates a historical live-stream sales diagram provided by an embodiment of the present application, such as... Figure 3-7 As shown, this was done to determine the sales records of five live streamers over 30 previous sessions.

[0126] By coupling the profiles of the five livestreamers with the products, a similarity matrix between the livestreamers and the products is obtained:

[0127] anchor Product 1 Product 2 Product 3 Product 4 Product 5 Product 6 Zheng Jiao 32923.8 14145.3 0 0 6453 18125 Su Huijun 40259.7 0 5211 38256 22167 0 Jin Xin 0 40786.2 30996 0 15593 0 Li Dongdong 0 48508.2 0 0 17546 0 Wang Ting 41347.8 51421.5 0 0 23940 0

[0128] By coupling the profiles of the five anchors, a similarity matrix among the five anchors is obtained and represented in a table:

[0129]

[0130]

[0131] The similarity matrix shows that anchor Su Huijun and anchor Wang Ting have a high similarity, and anchor Jin Xin and anchor Li Dongdong have a high similarity. Based on the anchor product rating matrix, product 2 is recommended to Su Huijun, product 4 to Wang Ting, product 3 to Li Dongdong, and product 5 to Jin Xin.

[0132] By inputting the above data into the product sales revenue prediction model, a list of product prediction scenarios is obtained:

[0133]

[0134] It can be seen that the optimized data from the 10 live streams showed significant improvements in both average profit per stream and average growth rate.

[0135] It should be noted that the various optional embodiments described in this application can be combined with each other or implemented individually without conflict, and this application does not limit the implementation of these embodiments.

[0136] Figure 8 A schematic diagram of the structure of a recommended device provided in an embodiment of this application is shown. For example... Figure 7As shown, the device includes the following modules:

[0137] The coupling calculation module 801 is used to perform coupling calculations on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of X merchants, and the first recommendation list corresponding to each merchant includes the identity information of the streamer whose matching value with the merchant is greater than a first threshold. Where the at least two types of profile sets include a streamer profile set, a product profile set, and a user profile set, the first list includes a second recommendation list corresponding to each of Y streamers, and the second recommendation list corresponding to each streamer includes product information corresponding to products in the streamer's recommended product set whose product recommendation value is greater than a second threshold. X and Y are positive integers.

[0138] The determining module 802 is used to determine the matching relationship between X merchants and Y live streamers based on the first list, wherein the matching relationship includes at least one of the following: each of the X merchants is matched with at least one live streamer; each of the Y live streamers is matched with at least one merchant;

[0139] The generation module 803 is used to generate a target recommendation list based on the matching relationship. Wherein, if the matching relationship includes at least one streamer matched by each of the X merchants, the target recommendation list includes the streamer recommendation list of each of the X merchants, and the streamer recommendation list of each merchant includes the identity information of at least one streamer matched by the merchant; if the matching relationship includes at least one merchant matched by each of the Y streamers, the target recommendation list includes the merchant recommendation list of each of the Y streamers, and the merchant recommendation list of each streamer includes the identity information and product information of at least one merchant matched by the streamer.

[0140] In the embodiments of this application, the coupling calculation module 801 performs coupling calculation on at least two types of portrait sets to obtain a first list, the determination module 802 determines the matching relationship between X merchants and Y live streamers based on the first list, and the generation module 803 generates a target recommendation list based on the matching relationship. Therefore, this application can couple calculations on at least two types of profile sets among the streamer profile set, product profile set, and user profile set to obtain a first recommendation list corresponding to each of the X merchants, so as to recommend streamers with a matching value greater than a first threshold to each merchant, and obtain a second recommendation list corresponding to each of the Y streamers, so as to recommend products in the streamer's recommended product set with a product recommendation value greater than the second threshold to each streamer. Based on the first list, the matching relationship between the X merchants and the Y streamers is determined to determine at least one streamer matched by each of the X merchants, or at least one merchant matched by each of the Y streamers. According to the matching relationship, a target recommendation list is generated. Streamers are recommended to each merchant through the streamer recommendation lists of the X merchants included in the target recommendation list, and merchants and their products are recommended to each streamer through the merchant recommendation lists of the Y streamers included in the target recommendation list. It is evident that the target recommendation list enables streamers to choose suitable products from among many merchants' products for promotion, and merchants to choose suitable streamers from among many streamers, thus improving recommendation capabilities.

[0141] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0142] The electronic device may include a processor 901 and a memory 902 storing computer program instructions.

[0143] Specifically, the processor 901 may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0144] Memory 902 may include mass storage for data or instructions. For example, and not limitingly, memory 902 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, memory 902 may include removable or non-removable (or fixed) media. Where appropriate, memory 902 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, memory 902 is non-volatile solid-state memory.

[0145] In certain embodiments, memory may include read-only memory (ROM), random access memory (RAM), disk storage media devices, optical storage media devices, flash memory devices, and electrical, optical, or other physical / tangible memory storage devices. Thus, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the methods according to one aspect of this disclosure.

[0146] The processor 901 implements any of the recommended methods in the above embodiments by reading and executing computer program instructions stored in the memory 902.

[0147] In one example, the electronic device may also include a communication interface 903 and a bus 910. Wherein, as... Figure 9 As shown, the processor 901, memory 902, and communication interface 903 are connected through bus 910 and complete communication with each other.

[0148] The communication interface 903 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0149] Bus 910 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 910 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0150] In addition, in conjunction with the recommended methods in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the recommended methods in the above embodiments.

[0151] This application provides a computer program product in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform the recommended method described above.

[0152] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0153] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0154] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0155] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0156] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A recommendation method, characterized in that, include: Coupled calculations are performed on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of X merchants, where each merchant's first recommendation list includes the streamer's identity information with a matching value greater than a first threshold. Where the at least two types of profile sets include the streamer profile set, the product profile set, and the user profile set, the first list includes a second recommendation list corresponding to each of Y streamers, where each streamer's second recommendation list includes product information corresponding to products in the streamer's recommended product set with a product recommendation value greater than a second threshold, where X and Y are positive integers. Based on the first list, the matching relationship between the X merchants and the Y live streamers is determined, wherein the matching relationship includes at least one of the following: each of the X merchants is matched with at least one live streamer; each of the Y live streamers is matched with at least one merchant; Based on the matching relationship, a target recommendation list is generated. Where the matching relationship includes at least one streamer matched by each of the X merchants, the target recommendation list includes a streamer recommendation list for each of the X merchants, and the streamer recommendation list for each merchant includes the identity information of at least one streamer matched by that merchant. Where the matching relationship includes at least one merchant matched by each of the Y streamers, the target recommendation list includes a merchant recommendation list for each of the Y streamers, and the merchant recommendation list for each streamer includes the identity information and product information of at least one merchant matched by that streamer. Wherein, when the first list is the first recommendation list, the step of performing coupled calculations on at least two types of image sets to obtain the first list includes: The coupling calculation is performed on the set of anchor profiles and the set of product profiles to obtain a second similarity matrix between anchors and products. Based on the products sold by the merchant, the feature vector of the merchant is obtained; Generate a second matrix, which includes the feature vectors of each merchant; Based on the second similarity matrix and the second matrix, a matching degree matrix between the merchant and the streamer is determined; Based on the matching degree matrix, target livestreamers with a matching degree greater than the seventh threshold are identified; Generate the first recommendation list, which includes the streamer information of the target streamer.

2. The method according to claim 1, characterized in that, When the first list is the second recommendation list, the step of performing coupled calculations on at least two types of image sets to obtain the first list includes: The coupling calculation is performed on each first object portrait in the portrait set corresponding to the first object to obtain a first set; wherein, when the first object is a user, the first set is a user recommended product set, which includes: Q users and the recommended products for purchase corresponding to the Q users; when the first object is a broadcaster, the first set is a broadcaster recommended product set, which includes: P broadcasters and the recommended products corresponding to the P broadcasters, where Q and P are positive integers. The coupling calculation is performed on the user profile set and the product profile set to obtain the first similarity matrix between users and products; Using the users and products in the first similarity matrix whose similarity is greater than a third threshold, a set of products that users are interested in is generated, wherein the set of products that users are interested in includes M users and the M products that the users are interested in, where M is a positive integer; The coupling calculation is performed on the user profile set and the anchor profile set to obtain the user's favorite anchor set, wherein the user's favorite anchor set includes N users and the N users' favorite anchors, where N is a positive integer; If there is an overlap between the products in the user-recommended product set and the products in the streamer-recommended product set of the streamer in the user-favorite streamer set, or if there is an overlap between the products in the user-interested product set and the products in the streamer-recommended product set of the streamer in the user-favorite streamer set, the product recommendation value of the product in the streamer-recommended product set will be increased by a preset value. If there is no overlap between the products in the user-recommended product set and the products in the streamer-recommended product set of the streamer in the user-favorite streamer set, the products in the user-recommended product set and the product recommendation values ​​of the products are added to the streamer-recommended product set. If there is no overlap between the products in the set of products that the user is interested in and the products in the set of products recommended by the streamers that the user likes, the products in the set of products that the user is interested in and the product recommendation values ​​of the products are added to the set of products recommended by the streamers. Obtain target products from the set of products recommended by the livestreamer whose product recommendation value is greater than the fourth threshold. Generate a second recommendation list, wherein the second recommendation column includes product information of the target product.

3. The method according to claim 2, characterized in that, The coupling calculation is performed on each first object image in the image set corresponding to the first object to obtain the first set, including: The coupling calculation is performed on the object profiles corresponding to each first object to determine the similarity between the first objects. Generate at least one first object pair, wherein each first object pair includes two first objects, and the similarity between the two first objects is greater than a fifth threshold; The first product lists of each first object in the first object pair are merged to obtain the product set of the first object pair. Wherein, when the first object is a user and the first product list is the user's historical purchase product list, the product set is the user's historical purchase product set. When the first object is a live streamer and the first product list is the live streamer's historical sales products, the product set is the live streamer's historical sales product set. Calculate the difference between the first product list of each first object in the first object pair and the product set to obtain the recommended product list of each first object; Generate the first set, which includes each first object and a list of recommended products corresponding to each first object, wherein the list of recommended products includes recommended products.

4. The method according to claim 3, characterized in that, The coupling calculation performed on the object profiles corresponding to each first object to determine the similarity between the first objects includes: Based on the first object information in the object profile corresponding to each first object, the feature vector of each first object is obtained; Generate a first column vector, which includes the feature vectors of each of the first objects; Based on the first column vector and its transpose, a first matrix is ​​determined. This first matrix comprises Z*Z elements, where the element in the i-th row and j-th column represents the similarity between the i-th and j-th first objects. Z, i, and j are positive integers. Z,j Z.

5. The method according to claim 2, characterized in that, The coupling calculation of the user profile set and the streamer profile set to obtain the user's favorite streamer set includes: The coupling calculation is performed on the user profile set and the anchor profile set to obtain the intimacy between each of the N users and the V anchors. Generate the set of favorite anchors of the users, which includes: the N users and the anchors among the N users whose intimacy with the V anchors is greater than a sixth threshold.

6. The method according to claim 2, characterized in that, The coupling calculation of the user profile set and the product profile set to obtain the first similarity matrix between users and products includes: Based on the second object information in the object profile corresponding to each second object, the feature vector of each second object is obtained, wherein the second object is the user; Generate a second column vector, which includes the feature vectors of each of the second objects; Based on the product feature information in the product profile corresponding to each product, the feature vector of each product is obtained; Generate a third column vector, which includes the feature vectors of each product; Based on the second column vector, its transpose, the third column vector, and its transpose, a target similarity matrix is ​​determined. This target similarity matrix comprises R*R elements and is the first similarity matrix. The element in the a-th row and b-th column represents the similarity between the a-th user and the b-th product. R, a, and b are positive integers. R, b R.

7. The method according to claim 1, characterized in that, The coupling calculation of the streamer profile set and the product profile set to obtain the second similarity matrix between the streamer and the product includes: Based on the information of the second object in the object profile corresponding to each second object, the feature vector of each second object is obtained, wherein the second object is the anchor; Generate a second column vector, which includes the feature vectors of each of the second objects; Based on the product feature information in the product profile corresponding to each product, the feature vector of each product is obtained; Generate a third column vector, which includes the feature vectors of each product; Based on the second column vector, its transpose, the third column vector, and its transpose, a target similarity matrix is ​​determined. This target similarity matrix comprises R*R elements and is the second similarity matrix. The element in the a-th row and b-th column represents the similarity between the a-th streamer and the b-th product. R, a, and b are positive integers. R, b R.

8. The method according to claim 1, characterized in that, The step of determining the matching relationship between the X merchants and the Y live streamers based on the first list includes: When the first list is the second recommendation list, obtain the distribution of merchant reviews in the product information of the second recommendation list corresponding to each of the Y streamers; Identify at least one merchant that matches each of the Y streamers, wherein the at least one merchant is one whose merchant rating distribution is greater than the ninth threshold.

9. The method according to claim 1, characterized in that, The step of determining the matching relationship between the X merchants and the Y live streamers based on the first list includes: If the first list is the first recommendation list, obtain at least one of the following from the anchor information in the first recommendation list corresponding to each of the X merchants: anchor monetization power or anchor charm value. Identify at least one streamer who matches each of the X merchants, wherein the at least one streamer is a streamer whose monetization ability or charm value is greater than the tenth threshold.

10. The method according to claim 1, characterized in that, After generating the target recommendation list based on the matching relationship, the method further includes: The target recommendation list is input into the product sales revenue prediction model to obtain the predicted average sales volume of the products in the target recommendation list, wherein the product sales revenue prediction model is trained using the historical data of the products; The difference between the predicted average sales volume and the average sales volume in the historical data is calculated to obtain the product's evaluation value; If the evaluation value is greater than or equal to the eleventh threshold, a formal recommendation list is generated.

11. The method according to claim 1, characterized in that, The streamer profile includes the streamer's monetization potential or streamer charm value, and the user profile includes the intimacy level between the user and each streamer. Before performing coupled calculations on at least two types of profiles to obtain the first list, the method further includes: The streamer's charm value is obtained by statistically analyzing and calculating the number of fans, the average number of users participating in each live stream, the average duration of user participation, the average number of likes received per live stream, the average number of tips received per live stream, and the average number of user interactions per live stream. The monetization power of the streamer is obtained by statistically analyzing and calculating the total sales, average sales per session, average number of orders per session, average number of link clicks per session, average number of orders per user per session, and average sales per user per session. The user's intimacy level with each streamer is determined by statistically analyzing and calculating the number of interactions with each streamer, the number of times the user tipped each streamer, the number of times the user liked each streamer's post, the average duration of participation in each streamer's live stream, the average number of orders placed during each live stream, and the average number of links clicked during each live stream.

12. A recommendation device, characterized in that, include: A coupling calculation module is used to perform coupling calculations on at least two types of profile sets to obtain a first list. Where the at least two types of profile sets include a streamer profile set and a product profile set, the first list includes a first recommendation list corresponding to each of X merchants, and the first recommendation list corresponding to each merchant includes the identity information of streamers whose matching value with the merchant is greater than a first threshold. Where the at least two types of profile sets include the streamer profile set, the product profile set, and the user profile set, the first list includes a second recommendation list corresponding to each of Y streamers, and the second recommendation list corresponding to each streamer includes product information corresponding to products in the streamer's recommended product set whose product recommendation value is greater than a second threshold, where X and Y are positive integers. The determining module is configured to determine the matching relationship between the X merchants and the Y live streamers based on the first list, wherein the matching relationship includes at least one of the following: at least one live streamer matched by each of the X merchants; at least one merchant matched by each of the Y live streamers; A generation module is used to generate a target recommendation list based on the matching relationship. Where the matching relationship includes at least one streamer matched by each of the X merchants, the target recommendation list includes a streamer recommendation list for each of the X merchants, and the streamer recommendation list for each merchant includes the identity information of at least one streamer matched by that merchant. Where the matching relationship includes at least one merchant matched by each of the Y streamers, the target recommendation list includes a merchant recommendation list for each of the Y streamers, and the merchant recommendation list for each streamer includes the identity information and product information of at least one merchant matched by that streamer. Wherein, when the first list is the first recommendation list, the coupling calculation module is specifically used for: The coupling calculation is performed on the set of anchor profiles and the set of product profiles to obtain a second similarity matrix between anchors and products. Based on the products sold by the merchant, the feature vector of the merchant is obtained; Generate a second matrix, which includes the feature vectors of each merchant; Based on the second similarity matrix and the second matrix, a matching degree matrix between the merchant and the streamer is determined; Based on the matching degree matrix, target livestreamers with a matching degree greater than the seventh threshold are identified; Generate the first recommendation list, which includes the streamer information of the target streamer.

13. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the recommended method as described in any one of claims 1-11.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the recommended method as described in any one of claims 1-11.

15. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device is able to perform the recommended method as described in any one of claims 1-11.

Images