Commercial tenant recommendation method of platform, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a platform merchant recommendation method, electronic equipment, a storage medium and a program product. The method comprises the steps of obtaining first historical behavior data when a current user interacts with a platform, and determining an initial recommendation basis of the current user according to the first historical behavior data; according to the first historical behavior data and a pre-constructed reference basis recommendation pool, determining a reference recommendation basis of the current user; wherein the reference basis recommendation pool is a data pool which is pre-constructed based on a large model according to second historical behavior data when a historical user interacts with the platform and is used for determining a reference recommendation basis of the current user; determining a final recommendation basis of the current user according to the initial recommendation basis and the reference recommendation basis; and according to the final recommendation basis, determining and presenting at least one to-be-recommended merchant. The method is used for achieving the effect of improving the accuracy and efficiency of merchant recommendation of the platform.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method for recommending merchants on a platform, an electronic device, a storage medium, and a program product. Background Art

[0002] During the use of the platform, the ways of reaching users are becoming more and more diverse. Therefore, how to accurately recommend the next step after the user is reached, and then introduce new users and promote activation, is an important issue currently faced.

[0003] In the prior art, the recommendation basis for the current user can be determined according to the historical behavior data of the current user, thereby determining the merchant to be recommended to the current user.

[0004] However, the methods in the prior art may lead to inaccurate recommendations. Summary of the invention

[0005] The embodiments of the present application provide a recommendation method, electronic device, storage medium and program product for merchants on the platform, so as to achieve the effect of improving the accuracy and efficiency of the recommendation of merchants on the platform.

[0006] In a first aspect, an embodiment of the present application provides a method for recommending merchants on a platform, comprising:

[0007] Acquire first historical behavior data when the current user interacts with the platform, and determine an initial recommendation basis for the current user based on the first historical behavior data;

[0008] Determine the reference recommendation basis of the current user according to the first historical behavior data and the pre-constructed reference basis recommendation pool; wherein the reference basis recommendation pool is a data pool pre-constructed based on the big model and the second historical behavior data when the historical user interacts with the platform, and is used to determine the reference recommendation basis of the current user;

[0009] According to the initial recommendation basis and the reference recommendation basis, a final recommendation basis for the current user is determined; and according to the final recommendation basis, at least one merchant to be recommended is determined and presented.

[0010] In a possible implementation, the process of constructing the reference recommendation basis pool includes: obtaining second historical behavior data of the historical user when interacting with the platform; and determining a reference recommendation basis pool based on the second historical behavior data based on a large model.

[0011] In a possible implementation, based on the big model, a reference recommendation basis pool is determined according to the second historical behavior data, including: determining a set of seed users from the historical users according to the second historical behavior data; wherein the seed user set includes at least one seed user; obtaining third historical behavior data of each seed user in the seed user set when interacting with the platform, and based on the big model, determining a reference recommendation basis for each seed user according to the third historical behavior data; and determining the reference basis recommendation pool according to the reference recommendation basis of each seed user.

[0012] In a possible implementation, based on the big model, according to the third historical behavior data, a reference recommendation basis for each seed user is determined, including: based on the third historical behavior data, determining the interaction behavior sequence of each seed user; based on the big model, determining the preference prompt words for each seed user according to the interaction behavior sequence of each seed user; based on the big model, determining the reference recommendation basis for each seed user according to the preference prompt words of each seed user.

[0013] In a possible implementation, determining a set of seed users from the historical users based on the second historical behavior data includes: determining a set of representative seed users from the historical users based on the second historical behavior data; determining a set of important seed users from the historical users based on the second historical behavior data; and determining the set of seed users based on the representative set of seed users and the set of important seed users.

[0014] In a possible implementation, a representative seed user set is determined from the historical users based on the second historical behavior data, including: based on a user behavior vector generation algorithm, processing the second historical behavior data to obtain a user behavior vector of each user; based on a clustering algorithm, clustering the user behavior vector of each user to obtain at least one user cluster set; the user cluster set includes at least one user; according to a preset extraction strategy, extracting at least one user from each user cluster set, determining the extracted user as the representative seed user, and determining a set consisting of representative seed users extracted from each user cluster set as the representative seed user set.

[0015] In a possible implementation, the preset extraction strategy includes: a preset ratio extraction strategy; extracting at least one user from each user cluster set according to the preset extraction strategy, including: for each user cluster set, determining the number of users included in the user cluster set; determining the number of users to be extracted in the user cluster set according to the extraction ratio indicated by the preset ratio extraction strategy and the number of users included in the user cluster set; extracting at least one user from the user cluster set according to the number of users to be extracted in the user cluster set.

[0016] In a possible implementation, the preset extraction strategy includes: a preset quantity extraction strategy; extracting at least one user from each user cluster set according to the preset extraction strategy, including: for each user cluster set, determining the cluster center point of the user cluster set; determining the distance between each user in the user cluster set and the cluster center point of the user cluster set; extracting at least one user from the user cluster set according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set.

[0017] In a possible implementation, at least one user is extracted from a user cluster set according to the extraction quantity indicated by the preset extraction quantity strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set, including: determining the number of users included in the user cluster set; if the extraction quantity indicated by the preset extraction quantity strategy is less than the number of users included in the user cluster set, extracting at least one user from the user cluster set according to the distance between each user in the user cluster set and the cluster center point; wherein the number of extracted users is equal to the extraction quantity indicated by the preset extraction quantity strategy; if the extraction quantity indicated by the preset extraction quantity strategy is greater than or equal to the number of users included in the user cluster set, extracting all users from the user cluster set.

[0018] In a possible implementation, based on the second historical behavior data, a set of important seed users is determined from the historical users, including: based on the second historical behavior data, determining the number of user interaction behaviors of each user; based on a preset sampling strategy and the number of user interaction behaviors of each user, determining the set of important seed users from the historical users and the current user.

[0019] In a possible implementation, the preset sampling strategy indicates a threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors; wherein the total number of user interaction behaviors is the sum of the number of user interaction behaviors of each user; according to the preset sampling strategy and the number of user interaction behaviors of each user, determining the set of importance seed users from the historical users, including: calculating the ratio of the number of user interaction behaviors of each user to the total number of user interaction behaviors; determining the user whose calculated ratio is greater than the threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors as the importance seed user in the set of importance seed users.

[0020] In a possible implementation, the preset sampling strategy also indicates a threshold value for the number of user interaction behaviors; before calculating the proportion of the number of user interaction behaviors of each user to the total number of user behaviors, the method further includes: screening users whose number of user interaction behaviors is greater than the threshold value for the number of user interaction behaviors.

[0021] In a possible implementation manner, a ratio of the number of the representative seed users included in the representative seed user set to the number of the important seed users included in the important seed user set is a preset ratio threshold.

[0022] In a possible implementation, the pre-constructed reference basis recommendation pool includes a reference recommendation basis for each seed user and a user behavior vector for each seed user; based on the first historical behavior data and the pre-constructed reference basis recommendation pool, determining the reference recommendation basis for the current user includes: based on a user behavior vector generation algorithm, processing the first historical behavior data to obtain a first user behavior vector for the current user; based on a vector similarity algorithm, calculating the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool; based on the calculated similarity, determining the reference recommendation basis for the current user from the reference recommendation basis for each seed user included in the pre-constructed reference recommendation basis pool.

[0023] In a possible implementation, the final recommendation basis for the current user is determined based on the initial recommendation basis and the reference recommendation basis, including: based on a vector generation algorithm, processing the initial recommendation basis and the reference recommendation basis to obtain an initial recommendation basis vector and a reference recommendation basis vector; performing vector fusion processing on the initial recommendation basis vector and the reference recommendation basis vector to obtain the final recommendation basis for the current user.

[0024] In a possible implementation, determining the initial recommendation basis for the current user based on the first historical behavior data includes: performing feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user; and determining the initial recommendation basis for the current user based on the behavior characteristics of the current user.

[0025] In a second aspect, an embodiment of the present application provides a recommendation device for merchants on a platform, including:

[0026] A first determination module, used to obtain first historical behavior data when the current user interacts with the platform, and determine the initial recommendation basis for the current user based on the first historical behavior data;

[0027] A second determination module is used to determine the reference recommendation basis of the current user based on the first historical behavior data and a pre-constructed reference basis recommendation pool; wherein the reference basis recommendation pool is a data pool pre-constructed based on a large model and the second historical behavior data when the historical user interacts with the platform, and is used to determine the reference recommendation basis of the current user;

[0028] The recommendation module is used to determine the final recommendation basis of the current user based on the initial recommendation basis and the reference recommendation basis; and determine and present at least one merchant to be recommended based on the final recommendation basis.

[0029] In a possible implementation, the device further includes a construction module for obtaining second historical behavior data of the historical user when interacting with the platform; based on the big model, a reference recommendation basis pool is determined according to the second historical behavior data.

[0030] In a possible implementation, the construction module is specifically used to determine a set of seed users from the historical users based on the second historical behavior data; wherein the seed user set includes at least one seed user; obtain the third historical behavior data of each seed user in the seed user set when interacting with the platform, and based on the big model, determine the reference recommendation basis for each seed user according to the third historical behavior data; determine the reference basis recommendation pool according to the reference recommendation basis of each seed user.

[0031] In a possible implementation, the construction module is further specifically used to determine the interaction behavior sequence of each seed user based on the third historical behavior data; based on the big model, determine the preference prompt words of each seed user according to the interaction behavior sequence of each seed user; based on the big model, determine the reference recommendation basis of each seed user according to the preference prompt words of each seed user.

[0032] In a possible implementation, the construction module is further specifically used to determine a representative seed user set from the historical users based on the second historical behavior data; determine an important seed user set from the historical users based on the second historical behavior data; and determine the seed user set based on the representative seed user set and the important seed user set.

[0033] In a possible implementation, the construction module is further specifically used to process the second historical behavior data based on a user behavior vector generation algorithm to obtain a user behavior vector for each user; cluster the user behavior vector for each user based on a clustering algorithm to obtain at least one user cluster set; the user cluster set includes at least one user; according to a preset extraction strategy, extract at least one user from each user cluster set, determine the extracted user as the representative seed user, and determine the set consisting of the representative seed users extracted from each user cluster set as the representative seed user set.

[0034] In a possible implementation, the preset extraction strategy includes: a preset ratio extraction strategy; the construction module is further specifically used to determine, for each user cluster set, the number of users included in the user cluster set; determine the number of users to be extracted from the user cluster set according to the extraction ratio indicated by the preset ratio extraction strategy and the number of users included in the user cluster set; and extract at least one user from the user cluster set according to the number of users to be extracted from the user cluster set.

[0035] In a possible implementation, the preset extraction strategy includes: a preset quantity extraction strategy; the construction module is further specifically used to determine the cluster center point of the user cluster set for each user cluster set; determine the distance between each user in the user cluster set and the cluster center point of the user cluster set; extract at least one user from the user cluster set according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set.

[0036] In a possible implementation, the construction module is further specifically used to determine the number of users included in the user cluster set; if the extraction number indicated by the preset number extraction strategy is less than the number of users included in the user cluster set, then at least one user is extracted from the user cluster set based on the distance between each user in the user cluster set and the cluster center point; wherein the number of extracted users is equal to the extraction number indicated by the preset number extraction strategy; if the extraction number indicated by the preset number extraction strategy is greater than or equal to the number of users included in the user cluster set, then all users are extracted from the user cluster set.

[0037] In a possible implementation, the construction module is further specifically used to determine the number of user interaction behaviors of each user based on the second historical behavior data; and determine the set of important seed users from the historical users and the current users based on a preset sampling strategy and the number of user interaction behaviors of each user.

[0038] In a possible implementation, the preset sampling strategy indicates a threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors; wherein the total number of user interaction behaviors is the sum of the number of user interaction behaviors of each user; the construction module is further specifically used to calculate the ratio of the number of user interaction behaviors of each user to the total number of user interaction behaviors; and the user whose calculated ratio is greater than the threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors is determined as an importance seed user in the importance seed user set.

[0039] In a possible implementation, the preset sampling strategy further indicates a threshold value for the number of user interaction behaviors; and the construction module is further specifically used to screen users whose number of user interaction behaviors is greater than the threshold value for the number of user interaction behaviors.

[0040] In a possible implementation manner, a ratio of the number of the representative seed users included in the representative seed user set to the number of the important seed users included in the important seed user set is a preset ratio threshold.

[0041] In a possible implementation, the pre-constructed reference basis recommendation pool includes a reference recommendation basis for each seed user and a user behavior vector for each seed user; the second determination module is specifically used to process the first historical behavior data based on a user behavior vector generation algorithm to obtain a first user behavior vector for the current user; based on a vector similarity algorithm, calculate the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool; and based on the calculated similarity, determine the reference recommendation basis for the current user from the reference recommendation basis for each seed user included in the pre-constructed reference recommendation basis pool.

[0042] In a possible implementation, the recommendation module is specifically used to process the initial recommendation basis and the reference recommendation basis based on a vector generation algorithm to obtain an initial recommendation basis vector and a reference recommendation basis vector; and perform vector fusion processing on the initial recommendation basis vector and the reference recommendation basis vector to obtain the final recommendation basis for the current user.

[0043] In a possible implementation, the first determination module is specifically configured to perform feature extraction processing on the first historical behavior data to obtain behavior features of the current user; and determine an initial recommendation basis for the current user based on the behavior features of the current user.

[0044] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory, a processor;

[0045] The memory stores computer-executable instructions;

[0046] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0047] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0048] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0049] The merchant recommendation method, electronic device, storage medium and program product of the platform provided in the embodiment of the present application obtain the first historical behavior data of the current user when interacting with the platform, and determine the initial recommendation basis of the current user according to the first historical behavior data, determine the reference recommendation basis of the current user according to the first historical behavior data and the pre-constructed reference basis recommendation pool, and determine the final recommendation basis of the current user according to the initial recommendation basis and the reference recommendation basis; and determine and present at least one merchant to be recommended according to the final recommendation basis, wherein, on the basis of the initial recommendation basis, the reference recommendation basis is optimized by obtaining the reference recommendation basis, thereby improving the accuracy of the reference recommendation basis and further improving the accuracy of the merchant recommendation of the platform, wherein the pre-constructed reference basis recommendation pool, based on the large model, introduces external knowledge, that is, the data of historical users other than the current user, enriches the representation of the object to be recommended, improves the user portrait, and compensates for the sparsity of the data, thereby improving the accuracy of the merchant recommendation of the platform; wherein, the reference recommendation basis is confirmed by the pre-constructed reference basis recommendation pool, the efficiency of the reference recommendation basis confirmation can be improved, thereby improving the efficiency of the merchant recommendation of the platform, and the above description is combined to improve the accuracy and efficiency of the merchant recommendation of the platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0051] Figure 1 A schematic diagram of an application scenario provided for this application;

[0052] Figure 2 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 1 ;

[0053] Figure 3 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 2 ;

[0054] Figure 4 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 3 ;

[0055] Figure 5 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 4 ;

[0056] Figure 6 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 5 ;

[0057] Figure 7A schematic diagram of the structure of the recommendation device for merchants on the platform provided by this application;

[0058] Figure 8 A schematic diagram of the structure of the electronic device provided in this application.

[0059] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0060] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0061] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0062] It should be noted that the merchant recommendation method, electronic device, storage medium and program product of the application platform can be used in the fields of data processing technology and finance, and can also be used in any field other than the fields of data processing technology and finance. The application field of the merchant recommendation method, electronic device, storage medium and program product of the application platform is not limited.

[0063] Figure 1 A schematic diagram of an application scenario provided by this application. Figure 1 As shown, the specific application scenario of the present application includes an electronic device 101 and a user 102. The user 102 triggers an instruction on the electronic device 101 to indicate the recommendation of a merchant, and the electronic device 101, in response to the trigger instruction of the user 102 on the interface, obtains the initial recommendation basis and the reference recommendation basis of the user 102, and determines and displays at least one merchant to be recommended on the interface of the electronic device 101 according to the determined initial recommendation basis and the reference recommendation basis.

[0064] In combination with the above scenario, it can be known that in the prior art, the recommendation basis of the current user can be determined according to the historical behavior data of the current user, thereby determining the merchant to be recommended to the current user. However, the method in the prior art will lead to the problem of inaccurate recommendation.

[0065] The merchant recommendation method of the platform provided by the present application obtains the first historical behavior data of the current user when interacting with the platform, and determines the initial recommendation basis of the current user based on the first historical behavior data, determines the reference recommendation basis of the current user based on the first historical behavior data and a pre-constructed reference basis recommendation pool, and determines the final recommendation basis of the current user based on the initial recommendation basis and the reference recommendation basis; and determines and presents at least one merchant to be recommended based on the final recommendation basis, wherein, on the basis of the initial recommendation basis, the reference recommendation basis is optimized by obtaining the reference recommendation basis, thereby improving the accuracy of the reference recommendation basis and further improving the accuracy of the merchant recommendation of the platform, wherein the pre-constructed reference basis recommendation pool, based on a large model, introduces external knowledge, namely, the data of historical users other than the current user, enriches the representation of the object to be recommended, improves the user portrait, and compensates for the sparsity of the data, thereby improving the accuracy of the merchant recommendation of the platform; wherein, the reference recommendation basis is confirmed by the pre-constructed reference basis recommendation pool, which can improve the efficiency of the reference recommendation basis confirmation, thereby improving the efficiency of the merchant recommendation of the platform, and based on the above description, improves the accuracy and efficiency of the merchant recommendation of the platform.

[0066] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0067] Figure 2 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 1 ,like Figure 2 As shown, the method includes:

[0068] Step S201: Acquire first historical behavior data when the current user interacts with the platform, and determine an initial recommendation basis for the current user based on the first historical behavior data.

[0069] Specifically, the first historical behavior data when the current user interacts with the platform can be obtained. The first historical behavior data is the data of the historical behavior of the current user interacting with the platform. The present application does not limit the process of obtaining the first historical behavior data when the current user interacts with the platform. Optionally, the historical usage data of the platform can be obtained first, and then the first historical behavior data when the current user interacts with the platform can be screened from the historical usage data of the platform.

[0070] Specifically, after obtaining the first historical behavior data when the current user interacts with the platform, the initial recommendation basis for the current user can be determined based on the first historical behavior data. The initial recommendation basis for the current user is the recommendation basis for determining the merchant to be recommended determined based on the first historical behavior data when the current user interacts with the platform.

[0071] Among them, this application does not limit the process of determining the initial recommendation basis of the current user based on the first historical behavior data. Any process that can determine the initial recommendation basis of the current user based on the first historical behavior data can be used as the process of determining the initial recommendation basis of the current user based on the first historical behavior data provided by this application. Optionally, feature extraction processing can be performed on the first historical behavior data to obtain the behavioral features of the current user. Then, the initial recommendation basis of the current user is determined based on the behavioral features of the current user.

[0072] Step S202: Determine the reference recommendation basis for the current user based on the first historical behavior data and the pre-built reference recommendation pool.

[0073] Specifically, the reference recommendation basis of the current user can be determined based on the first historical behavior data and the pre-built reference basis recommendation pool. The description of the first historical behavior data can refer to the description in step S201, which will not be repeated here. The reference recommendation basis is determined based on the second historical behavior data and is used as a reference in the process of determining the user to be recommended.

[0074] The reference recommendation pool is a data pool pre-built based on a large model and second historical behavior data of historical users interacting with the platform, and is used to determine the reference recommendation basis for the current user;

[0075] The historical user is at least one user among all users interacting with the platform, other than the current user. The second historical behavior data is data on historical behaviors of the historical user interacting with the platform.

[0076] Among them, this application does not limit the reference basis recommendation pool. Any data pool based on the big model and pre-constructed according to the second historical behavior data when the historical user interacts with the platform and used to determine the reference recommendation basis of the current user can be used as the reference basis recommendation pool provided by this application. Optionally, the construction process of the reference basis recommendation pool may include: first obtaining the second historical behavior data when the historical user interacts with the platform; then based on the big model, according to the second historical behavior data, determine the reference recommendation pool.

[0077] Optionally, the pre-constructed reference basis recommendation pool can be dynamically maintained, wherein the present application does not limit the process of dynamically maintaining the pre-constructed reference basis recommendation pool. Optionally, the pre-constructed reference basis recommendation pool can be dynamically maintained in a preset maintenance cycle. For example, if the process of pre-constructing the reference basis recommendation pool is as described above, the second historical behavior data of historical users interacting with the platform can be re-obtained in the preset maintenance cycle; then based on the big model, the reference recommendation basis pool is re-determined according to the re-obtained second historical behavior data.

[0078] Among them, dynamic maintenance of the pre-built reference basis recommendation pool can improve the real-time and accuracy of the reference basis recommendation pool, and further improve the accuracy of the platform's merchant recommendations.

[0079] Among them, the present application does not limit the process of determining the reference recommendation basis of the current user based on the first historical behavior data and the pre-constructed reference basis recommendation pool. Optionally, if the reference recommendation basis pool pre-constructed according to the above process includes the reference recommendation basis of each seed user and the user behavior vector of each seed user, then the process of determining the reference recommendation basis of the current user based on the first historical behavior data and the pre-constructed reference basis recommendation pool may include: firstly, based on the user behavior vector generation algorithm, the first historical behavior data is processed to obtain the first user behavior vector of the current user; then, based on the vector similarity algorithm, the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool is calculated; then, based on the calculated similarity, the reference recommendation basis of the current user is determined from the reference recommendation basis of each seed user included in the pre-constructed reference recommendation basis pool. Step S203: Determine the final recommendation basis of the current user based on the initial recommendation basis and the reference recommendation basis; and determine and present at least one merchant to be recommended based on the final recommendation basis.

[0080] Specifically, according to the initial recommendation basis obtained in step S201 and the reference recommendation basis obtained in step S202, the final recommendation basis for the current user can be determined. The final recommendation basis is determined based on the initial recommendation basis and the reference recommendation basis, and is used to determine the final recommendation basis for the user to be recommended. The present application does not limit the process of determining the final recommendation basis for the user to be recommended based on the initial recommendation basis and the reference recommendation basis. Optionally, the initial recommendation basis and the reference recommendation basis can be processed based on a vector generation algorithm to obtain an initial recommendation basis vector and a reference recommendation basis vector; then the initial recommendation basis vector and the reference recommendation basis vector are subjected to vector fusion processing to obtain the final recommendation basis for the current user.

[0081] Specifically, after determining the final recommendation basis of the current user based on the initial recommendation basis and the reference recommendation basis, at least one merchant to be recommended can be determined and presented based on the final recommendation basis. Among them, the present application does not limit the process of determining and presenting at least one merchant to be recommended based on the final recommendation basis. Any process that can determine and present at least one merchant to be recommended based on the final recommendation basis can be used as the process of determining and presenting at least one merchant to be recommended based on the final recommendation basis provided by the present application. Optionally, the degree of match between the current user and each merchant to be recommended on the platform can be determined based on the final recommendation basis, and then at least one merchant to be recommended can be determined and presented based on the determined degree of match between the current user and each merchant to be recommended on the platform. Among them, the present application does not limit the process of determining and presenting at least one merchant to be recommended based on the determined degree of match between the current user and each merchant to be recommended in the platform. Optionally, if multiple merchants to be recommended are determined and presented based on the determined degree of match between the current user and each merchant to be recommended in the platform, then the multiple merchants to be recommended are presented in the form of a list to be recommended, wherein the arrangement order of each merchant to be recommended in the list to be recommended corresponds to the degree of match between the current user and each merchant to be recommended.

[0082] The merchant recommendation method of the platform provided in the embodiment of the present application obtains the first historical behavior data of the current user when interacting with the platform, and determines the initial recommendation basis of the current user according to the first historical behavior data, determines the reference recommendation basis of the current user according to the first historical behavior data and a pre-constructed reference basis recommendation pool, and determines the final recommendation basis of the current user according to the initial recommendation basis and the reference recommendation basis; and determines and presents at least one merchant to be recommended according to the final recommendation basis, wherein, on the basis of the initial recommendation basis, the reference recommendation basis is optimized by obtaining the reference recommendation basis, thereby improving the accuracy of the reference recommendation basis and further improving the accuracy of the merchant recommendation of the platform, wherein the pre-constructed reference basis recommendation pool, based on a large model, introduces external knowledge, that is, the data of historical users other than the current user, enriches the representation of the object to be recommended, improves the user portrait, and compensates for the sparsity of the data, thereby improving the accuracy of the merchant recommendation of the platform; wherein, the reference recommendation basis is confirmed by the pre-constructed reference basis recommendation pool, which can improve the efficiency of the reference recommendation basis confirmation, thereby improving the efficiency of the merchant recommendation of the platform, and in combination with the above description, improves the accuracy and efficiency of the merchant recommendation of the platform.

[0083] Figure 3 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 2 ,like Figure 3 As shown, in this embodiment Figure 2 Based on the illustrated embodiment, the construction process of the recommendation pool is described in detail with reference to the method comprising:

[0084] Step S301: Acquire second historical behavior data of historical users interacting with the platform.

[0085] Specifically, the second historical behavior data when the historical user interacts with the platform can be obtained. For the description of the historical user and the second historical data, reference can be made to the description in step S202, which will not be repeated here.

[0086] Among them, this application does not limit the process of obtaining the second historical behavior data when the historical user interacts with the platform. Optionally, all the historical usage data of the platform can be obtained first, and then the second historical behavior data when the historical user interacts with the platform can be screened from all the historical usage data of the platform. Among them, this application does not limit the method of screening the second historical behavior data when the historical user interacts with the platform from all the historical usage data of the platform. Optionally, it can be screened according to preset screening conditions such as a preset time range.

[0087] Step S302: Based on the large model and the second historical behavior data, a reference recommendation basis pool is determined.

[0088] Specifically, after obtaining the second historical behavior data of the historical user interacting with the platform, a reference recommendation basis pool can be determined based on the big model and the second historical behavior data. Among them, this application does not limit the process of determining the reference recommendation basis pool based on the big model and the second historical behavior data. Optionally, the process of determining the reference recommendation basis pool based on the big model and the second historical behavior data may include:

[0089] First, according to the second historical behavior data, a seed user set is determined from historical users; wherein the seed user set includes at least one seed user;

[0090] Specifically, the seed user set includes at least one seed user. The seed user is selected from the historical users and is used to determine the user of the reference recommendation basis pool. The present application does not limit the process of determining the seed user set from the historical users according to the second historical behavior data. Optionally, a representative seed user set in the seed user set can be determined from the historical users according to the second historical behavior data, an important seed user set can be determined from the historical users according to the second historical behavior data, and a seed user set can be determined according to the representative seed user set and the important seed user set. The present application does not limit the number of seed users included in the determined seed user set, and the number of seed users included in the determined seed user set corresponds to the number of historical users corresponding to the second historical behavior data and the process of determining the seed user set from the historical users according to the second historical behavior data. That is, if the number of historical users corresponding to the second historical behavior data is different, and / or the process of determining the seed user set from the historical users according to the second historical behavior data is different, the number of seed users included in the determined seed user set is also different.

[0091] Secondly, obtain the third historical behavior data of each seed user in the seed user set when interacting with the platform, and determine the reference recommendation basis for each seed user based on the big model and the third historical behavior data;

[0092] The third historical behavior data is the data of the historical behavior of the seed user interacting with the platform. The present application does not limit the process of obtaining the third historical behavior data of each seed user in the seed user set when interacting with the platform. Optionally, the historical behavior data of each seed user interacting with the platform can be obtained from the second historical behavior data according to the seed user set.

[0093] Specifically, after obtaining the third historical behavior data of each seed user in the seed user set when interacting with the platform, the reference recommendation basis for each seed user can be determined based on the big model and the third historical behavior data. Among them, the present application does not limit the process of determining the reference recommendation basis for each seed user based on the big model and the third historical behavior data. Optionally, the interaction behavior sequence of each seed user can be determined based on the third historical behavior data, and the preference prompt words of each seed user can be generated based on the big model according to the interaction behavior sequence of each seed user and the preset prompt word generation strategy, and the reference recommendation basis of each seed user can be determined based on the big model according to the preference prompt words of each seed user.

[0094] Then, according to the reference recommendation basis of each sub-user, a reference basis recommendation pool is determined.

[0095] Specifically, after determining the reference recommendation basis for each seed user based on the large model and the third historical behavior data, a reference basis recommendation pool can be determined based on the reference recommendation basis for each seed user. Among them, the present application does not limit the process of determining the reference basis recommendation pool based on the reference recommendation basis for each seed user. Optionally, the user behavior vector of each seed user can be obtained, and then the user behavior vector of each seed user and the reference recommendation basis for each seed user can be combined, and the combined data corresponding to each seed user can be determined as the reference basis recommendation pool. Optionally, the reference basis recommendation pool includes the reference recommendation basis for each seed user and the user behavior vector of each seed user. For example, if the set of seed users is S, the reference recommendation basis set of the seed user set is , the user behavior vector set of seed users is M, then the formula of the reference recommendation pool P is as follows:

[0096]

[0097] in, is the user behavior vector of seed user s, It is a reference recommendation basis for seed users.

[0098] Among them, in the process of determining the reference recommendation basis pool based on the second historical behavior data based on the big model described above, the second historical behavior data is screened by determining a collection of seed users, thereby improving the reliability of the introduced external knowledge and further improving the accuracy of the constructed reference basis recommendation pool, thereby improving the accuracy of recommendations for merchants on the platform.

[0099] The process of constructing the reference basis recommendation pool provided in the embodiment of the present application obtains the second historical behavior data of historical users when interacting with the platform, and determines the reference recommendation basis pool based on the big model and the second historical behavior data. The second historical behavior data is used to introduce external knowledge, enrich the representation of the recommended objects, improve the user portrait, and compensate for the sparsity of the data, thereby improving the accuracy of the platform's merchant recommendations. The construction of the reference basis recommendation pool based on the big model improves the efficiency and accuracy of the construction.

[0100] Figure 4 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 3 ,like Figure 4 As shown, in this embodiment Figure 2 or Figure 3 On the basis of the illustrated embodiment, a process of determining a reference recommendation basis for each seed user based on a large model and according to the third historical behavior data is described in detail. The method includes:

[0101] Step S401: Determine the interaction behavior sequence of each sub-user according to the third historical behavior data.

[0102] Specifically, the interaction behavior sequence of each sub-user can be determined according to the third historical behavior data. For the description of the third historical behavior data, reference can be made to the description in step S302, which will not be repeated here. The interaction behavior sequence is a sequence obtained by extracting the interaction behaviors in the third historical behavior data.

[0103] The present application does not limit the process of determining the interaction behavior sequence of each seed user based on the third historical behavior data. Optionally, the interaction behavior sequence of each seed user can be determined based on steps such as filtering noise text and extracting interaction behaviors. For example, an example of an interaction behavior sequence provided in this embodiment can be: ={diapers, milk powder, dress, cap, backpack}.

[0104] Step S402: Based on the large model and according to the interaction behavior sequence of each sub-user, determine the preferred prompt words of each sub-user.

[0105] Specifically, based on the large model, according to the interaction behavior sequence of each seed user determined in step S401, the preferred prompt word of each seed user can be determined, wherein the preferred prompt word of the seed user is used to indicate the reference recommendation basis for generating the seed user.

[0106] Among them, the present application does not limit the process of determining the preference prompt words of each seed user based on the interaction behavior sequence of each seed user. Optionally, the interaction behavior sequence of each seed user can be input into a preset prompt word generation big model, for example, a big model for generating preference prompt words pre-trained based on the GLM4 big model can output the preference prompt words of each seed user.

[0107] Optionally, the preferred prompt words of each sub-user may be generated according to the preset prompt word template and the interaction behavior sequence of each sub-user. The present application does not limit the preset prompt word template, and any template that can be combined with the interaction behavior sequence of each sub-user to generate the preferred prompt words of each sub-user may be used as the preset prompt word template provided by the present application.

[0108] For example, a preset prompt word template example provided in this embodiment may be:

[0109] Based on the following constraints, extract keywords of user preferences based on the user's behavior sequence, such as "technology enthusiast", "fashionable and simple style", "high-end quality pursuer", "healthy life advocate", etc., and generate 3-5 keywords:

[0110] 1. Product categories: First, identify the main categories of products purchased by users, such as electronic products, fashion apparel, household items, books and education, etc.

[0111] 2. Brand preference: Observe whether users prefer a specific brand, which often reflects the user's brand loyalty or recognition of the quality of a specific brand.

[0112] 3. Price range: Pay attention to the price range in which the products purchased by users are mainly concentrated. This will help understand the user’s spending power and perception of value.

[0113] 4. Style and color: For categories such as fashion and home, analyze users’ preferred design styles (such as simple, retro, and luxurious) and color preferences (such as neutral colors and bright colors).

[0114] 5. Functional requirements: For functional products, such as electronic products or fitness equipment, pay attention to whether the products selected by users have certain specific functions, which reflects the actual needs and usage scenarios of users.

[0115] 6. Repeat purchasing behavior: Whether a user makes multiple purchases of a certain type of product is not only a reflection of satisfaction, but also a sign of strong preference.

[0116] 7. Related purchases: Analyze the correlation between the products purchased by users. For example, if a camera and lens are purchased at the same time, it may indicate that the user has a strong interest in photography.

[0117] 8. Reviews and Feedback: Check users’ reviews and feedback on purchased products, especially the advantages mentioned in positive reviews.

[0118] Step S403: Based on the large model, and according to the preference prompt words of each seed user, a reference recommendation basis for each seed user is determined.

[0119] Specifically, based on the large model, according to the preference prompt words of each seed user determined in step S402, a reference recommendation basis for each seed user may be determined.

[0120] Among them, the present application does not limit the process of determining the reference recommendation basis for each seed user based on the big model and the preference prompt words of each seed user. Optionally, the preference prompt words of each seed user can be input into a pre-built recommendation basis generation model, for example: a big model for generating recommendation basis pre-trained based on the GLM4 big model outputs the reference recommendation basis for each seed user.

[0121] For example, the interaction behavior sequence of the seed user shown in step S401 and the preference prompt words of the seed user generated in step S402 are input into the pre-built recommendation basis generation model, and the corresponding reference recommendation basis can be output: "User The keywords are: family and parenting, fashion, and pragmatists. ”

[0122] The embodiment of the present application provides a process for determining a reference recommendation basis for each seed user based on the third historical behavior data, by determining the interaction behavior sequence of each seed user based on the third historical behavior data, determining the preference prompt words of each seed user based on the big model and the reference recommendation basis of each seed user based on the preference prompt words of each seed user, wherein the interaction behavior sequence of the seed user is extracted from the third historical behavior data, which can better reflect the interaction behavior characteristics of the seed user, thereby improving the accuracy of determining the reference recommendation basis, wherein, after determining the interaction behavior sequence of the seed user, generating the reference recommendation basis by generating user preference prompt words based on the big model can improve the efficiency and accuracy of generating the reference recommendation basis, and further improve the accuracy and efficiency of the recommendation of the merchants of the platform.

[0123] Figure 5 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 4 ,like Figure 5 As shown, in this embodiment Figure 2 or Figure 3 or Figure 4 Based on the illustrated embodiment, a process of determining a set of seed users from historical users according to the second historical behavior data is described in detail. The method includes:

[0124] Step S501: Determine a representative seed user set from historical users according to the second historical behavior data.

[0125] The representative seed user set includes at least one representative seed user. The important seed user set includes at least one important seed user. The representative seed user refers to a user whose user behavior data represents the representative behavior of the user. The important seed user refers to a user whose user behavior represents the importance of the user's behavior, for example, a user with a large number of user interaction behaviors.

[0126] Specifically, a representative set of seed users in the set of seed users may be determined from the historical users according to the second historical behavior data. The present application does not limit the process of determining a representative set of seed users in the set of seed users from the historical users according to the second historical behavior data. Optionally, the process of determining a representative set of seed users in the set of seed users from the historical users according to the second historical behavior data may include:

[0127] Firstly, based on a user behavior vector generation algorithm, the second historical behavior data is processed to obtain a user behavior vector for each user.

[0128] Among them, this application does not limit the process of processing the second historical behavior data based on the user behavior vector generation algorithm to obtain the user behavior vector of each user. Optionally, a lightweight behavior encoder (LightGCN) can be used to generate and encode the second historical behavior data to obtain the behavior representation vector of each user: .

[0129] Secondly, based on the clustering algorithm, the user behavior vector of each user is clustered to obtain at least one user clustering set.

[0130] The present application does not limit the process of clustering the user behavior vector of each user based on the clustering algorithm to obtain at least one user clustering set. Optionally, the current user and historical users can be clustered into K clusters based on the Kmeans clustering algorithm. The user clustering set includes at least one user.

[0131] Then, according to a preset extraction strategy, at least one user is extracted from each user cluster set, the extracted user is determined as a representative seed user, and a set consisting of representative seed users extracted from each user cluster set is determined as a representative seed user set.

[0132] Among them, the present application does not limit the process of extracting at least one user from each user cluster set according to the preset extraction strategy and determining the extracted user as a representative seed user. Optionally, the process of extracting at least one user from each user cluster set according to the preset extraction strategy and determining the extracted user as a representative seed user corresponds to the preset extraction strategy. Among them, the present application does not limit the preset extraction strategy, and any extraction strategy can be used as the preset extraction strategy provided by the present application. Optionally, the preset extraction strategy can be a preset ratio extraction strategy. Optionally, the preset extraction strategy can also be a preset quantity extraction strategy.

[0133] Specifically, after at least one user is extracted from each user cluster set according to a preset extraction strategy and the extracted user is determined as a representative seed user, a set consisting of representative seed users extracted from each user cluster set can be determined as a representative seed user set. .

[0134] Optionally, the preset extraction strategy may include: a preset ratio extraction strategy, wherein the preset ratio extraction strategy indicates an extraction ratio, wherein the extraction ratio is a ratio of the users to be extracted from the user cluster set to all users in the user cluster set.

[0135] Optionally, if the preset extraction strategy includes: a preset ratio extraction strategy, then the process of extracting at least one user from each user cluster set according to the preset extraction strategy may include:

[0136] First, for each user cluster set, determine the number of users included in the user cluster set. For example: for the kth cluster , including the number of users ; Secondly, according to the extraction ratio indicated by the preset ratio extraction strategy and the number of users included in the user cluster set, the number of users to be extracted from the user cluster set is determined. For example: if the extraction ratio indicated by the preset ratio extraction strategy is 1:2, then the number of users to be extracted is ; Then, according to the number of users to be extracted in the user cluster set, at least one user is extracted from the user cluster set. The present application does not limit the process of extracting at least one user from the user cluster set according to the number of users to be extracted in the user cluster set. Optionally, at least one user can be extracted from the user cluster set in a random manner according to the number of users to be extracted in the user cluster set. If the preset extraction strategy includes: a preset ratio extraction strategy, in the process of extracting at least one user from each user cluster set according to the preset extraction strategy, if the calculated number of users to be extracted is not an integer, it can be rounded up or rounded down.

[0137] Optionally, the preset extraction strategy may include: a preset number extraction strategy, wherein the preset number extraction strategy indicates the extraction number, wherein the extraction number is the number of users to be extracted from the user cluster set.

[0138] Optionally, if the preset extraction strategy includes: a preset number extraction strategy, then the process of extracting at least one user from each user cluster set according to the preset extraction strategy may include:

[0139] First, for each user cluster set, determine the cluster center of the user cluster set. Secondly, determine the distance between each user in the user cluster set and the cluster center of the user cluster set. Among them, the present application does not limit the process of determining the cluster center of the user cluster set and determining the distance between each user in the user cluster set and the cluster center of the user cluster set. Any process that can determine the cluster center of the user cluster set and determine the distance between each user in the user cluster set and the cluster center of the user cluster set can be used as the process of determining the cluster center of the user cluster set and determining the distance between each user in the user cluster set and the cluster center of the user cluster set provided by the present application. Then, according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center of the user cluster set, extract at least one user from the user cluster set.

[0140] The present application does not limit the process of extracting at least one user from the user cluster set according to the extraction quantity indicated by the preset extraction strategy and the distance between each user in the user cluster set and the cluster center of the user cluster set. Optionally, the process of extracting at least one user from the user cluster set according to the extraction quantity indicated by the preset extraction strategy and the distance between each user in the user cluster set and the cluster center of the user cluster set may include: determining the number of users included in the user cluster set. If the extraction quantity indicated by the preset extraction strategy is less than the number of users included in the user cluster set, at least one user is extracted from the user cluster set according to the distance between each user in the user cluster set and the cluster center. Among them, the present application does not limit the process of extracting at least one user from the user cluster set according to the distance between each user in the user cluster set and the cluster center point. Optionally, according to the distance between each user in the user cluster set and the cluster center point, the user with the smallest distance from the cluster center point can be extracted from the user cluster set in sequence until the number of users to be extracted is reached. For example: if the cluster set includes two users, and the number of extractions indicated by the preset number extraction strategy is one, then the user with a closer direct distance to the cluster center among the two users included in the cluster set is extracted as a representative seed user. Among them, the number of users extracted is equal to the number of extractions indicated by the preset number extraction strategy. Among them, if the number of extractions indicated by the preset number extraction strategy is greater than or equal to the number of users included in the user cluster set, all users are extracted from the user cluster set. For example: if the cluster set includes one user, and the number of extractions indicated by the preset number extraction strategy is two, then one user included in the cluster set is extracted as a representative seed user.

[0141] Among them, the set of representative seed users obtained in the above manner can cover a wider range of hobbies and behavior patterns of all users, ensuring the diversity of seed users.

[0142] Step S502: Determine an important seed user set from historical users according to the second historical behavior data.

[0143] Specifically, the important seed user set can be determined from the historical users according to the second historical behavior data. The present application does not limit the process of determining the important seed user set from the historical users according to the second historical behavior data. The process of determining the important seed user set from the historical users according to the second historical behavior data may include:

[0144] First, the number of user interaction behaviors of each user is determined based on the second historical behavior data. For example, the set of historical users is U, represents the set of interaction behaviors of user u, is the number of user interaction behaviors of user u.

[0145] Secondly, according to the preset sampling strategy and the number of user interaction behaviors of each user, a set of important seed users is determined from historical users.

[0146] The present application does not limit the process of determining the important seed user set from historical users according to the preset sampling strategy and the number of user interaction behaviors of each user. Optionally, the preset sampling strategy may indicate a threshold value of the ratio of the number of user interaction behaviors to the number of all user interaction behaviors. The number of all user interaction behaviors is the sum of the number of user interaction behaviors of each user.

[0147] Specifically, if a sampling strategy is preset, indicating a threshold value of the ratio of the number of user interaction behaviors to the number of all user interaction behaviors, then according to the preset sampling strategy and the number of user interaction behaviors of each user, the process of determining the important seed user set from historical users may include:

[0148] First, calculate the ratio of each user's user interaction behavior to the total number of user interaction behaviors. For example, if the set of historical users is U, represents the set of interaction behaviors of user u, is the number of user interaction behaviors of user u, and the calculation formula for the ratio of the number of user interaction behaviors of each user to the total number of user interaction behaviors is as follows:

[0149]

[0150] Then, the users whose calculated ratio is greater than the threshold of the ratio of the number of user interaction behaviors to the number of all user interaction behaviors are determined as the important seed users in the important seed user set. For example, if the set of historical users is U, represents the set of interaction behaviors of user u, is the number of user interaction behaviors of user u, then the set of important seed users The formula is as follows:

[0151]

[0152] in, is an indicator function, if The sample is an important seed user.

[0153] Optionally, based on the preset sampling strategy indicating the threshold of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors, the preset sampling strategy may also indicate a threshold of the number of user interaction behaviors. The threshold of the number of user interaction behaviors is the threshold of the number of user interaction behaviors included in the important seed users, that is, the minimum number of user interaction behaviors required to be included in the important seed users.

[0154] Optionally, before calculating the proportion of the number of user interaction behaviors of each user to all user behaviors, the process of determining a set of important seed users from historical users based on a preset sampling strategy and the number of user interaction behaviors of each user may also include: screening users whose number of user interaction behaviors is greater than a user interaction behavior threshold.

[0155] Among them, the set of important seed users obtained in the above manner retains more reliable and information-rich important users.

[0156] Step S503: Determine a seed user set according to the representative seed user set and the important seed user set.

[0157] The present application does not limit the process of determining the seed user set based on the representative seed user set and the important seed user set. Optionally, the union of the representative seed user set and the important seed user set can be determined as the seed user set. , the set of important seed users is , then the formula for the seed user set is as follows:

[0158]

[0159] Optionally, the ratio of the number of representative seed users included in the representative seed user set to the number of important seed users included in the important seed user set is a preset ratio threshold. Wherein, by balancing the number of representative seed users and the number of important seed users in the seed user set through the preset ratio threshold, the quality of the seed user set can be improved, and the accuracy of the recommendation of merchants on the platform based on the seed user set can be further improved. Wherein, the present application does not limit the preset ratio threshold, and any ratio that can balance the number of representative seed users and the number of important seed users in the seed user set to improve the quality of the seed user set can be used as the preset ratio threshold provided by the present application, for example: a 1:1 ratio.

[0160] The embodiment of the present application provides a process for determining a set of seed users from historical users based on the second historical behavior data, by determining a representative set of seed users in the set of seed users from historical users based on the second historical behavior data, determining an important set of seed users from historical users based on the second historical behavior data, and determining a set of seed users based on the representative set of seed users and the important set of seed users, wherein the seed user set is differentiated into a representative set of seed users and an important set of seed users, which can retain more reliable and information-rich important users on the basis of ensuring the representativeness of the seed users, thereby improving the quality of the seed user set and further improving the accuracy of recommending merchants on the platform based on the seed user set.

[0161] Figure 6 Schematic diagram of the recommendation method for merchants on the platform provided by this application Figure 5 ,like Figure 6 As shown, in this embodiment Figure 2 or Figure 3 or Figure 4 or Figure 5 Based on the illustrated embodiment, a process of determining a reference recommendation basis for the current user based on the first historical behavior data and a pre-built reference recommendation pool is described in detail. The method includes:

[0162] Step S601: Based on a user behavior vector generation algorithm, the first historical behavior data is processed to obtain a first user behavior vector of the current user.

[0163] The first user behavior vector is a vector obtained by vectorizing the first behavior data. The present application does not limit the process of processing the first historical behavior data based on the user behavior vector generation algorithm to obtain the first user behavior vector of the current user. Optionally, the first behavior data can be input into a lightweight behavior encoder to output the first user behavior vector.

[0164] The reference recommendation pool includes the reference recommendation basis of each seed user and the user behavior vector of each seed user.

[0165] Step S602: Calculate the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool based on a vector similarity algorithm.

[0166] Among them, the present application does not limit the process of calculating the similarity between the first user behavior vector and the user behavior vector of each seed user based on the vector similarity algorithm. Any process that can calculate the similarity between the first user behavior vector and the user behavior vector of each seed user based on the vector similarity algorithm can be used as the process of calculating the similarity between the first user behavior vector and the user behavior vector of each seed user based on the vector similarity algorithm provided by the present application. Among them, the calculation formula of the similarity between the first user behavior vector and the user behavior vector of each seed user is as follows:

[0167]

[0168] in, is the first user behavior vector, is the user behavior vector of the seed user.

[0169] Step S603: Determine the reference recommendation basis for the current user from the reference recommendation basis for each sub-user included in the pre-constructed reference recommendation basis pool according to the calculated similarity.

[0170] Specifically, after calculating the similarity between the first user behavior vector and the user behavior vector of each seed user based on the vector similarity algorithm, the reference recommendation basis of the current user can be determined from the reference recommendation basis of each seed user included in the reference recommendation basis pool according to the calculated similarity. Among them, the present application does not limit the process of determining the reference recommendation basis of the current user from the reference recommendation basis of each seed user included in the reference recommendation basis pool according to the calculated similarity. Optionally, the reference recommendation basis of the seed user corresponding to the user behavior vector with the highest similarity to the first user behavior vector can be determined as the reference recommendation basis of the current user.

[0171] The formula for the reference recommendation of the current user is as follows:

[0172]

[0173] in, is the reference recommendation basis for the current user, and S is the set of seed users.

[0174] The embodiment of the present application provides a process for determining a reference recommendation basis for a current user based on first historical behavior data and a pre-constructed reference basis recommendation pool. The process processes the first historical behavior data based on a user behavior vector generation algorithm to obtain a first user behavior vector of the current user. Based on a vector similarity algorithm, the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool is calculated. Based on the calculated similarity, the reference recommendation basis for the current user is determined from the reference recommendation basis for each seed user included in the pre-constructed reference recommendation basis pool. Based on the vector similarity algorithm, the reference recommendation basis for the current user can be efficiently and accurately determined from the pre-constructed reference basis recommendation pool, thereby improving the efficiency and accuracy of the platform's merchant recommendations.

[0175] In a possible embodiment, the process of determining the final recommendation basis for the current user according to the initial recommendation basis and the reference recommendation basis may include:

[0176] First, based on the vector generation algorithm, the initial recommendation basis and the reference recommendation basis are processed to obtain the initial recommendation basis vector and the reference recommendation basis vector. Specifically, after determining the initial recommendation basis and the reference recommendation basis of the current user, the initial recommendation basis and the reference recommendation basis can be vectorized based on the vector generation algorithm to obtain the initial recommendation basis vector and the reference recommendation basis vector.

[0177] Secondly, the initial recommendation basis vector and the reference recommendation basis vector are fused to obtain the final recommendation basis for the current user.

[0178] Specifically, after obtaining the initial recommendation basis vector and the reference recommendation basis vector, the initial recommendation basis vector and the reference recommendation basis vector can be subjected to vector fusion processing to obtain the final recommendation basis for the current user. Among them, the present application does not limit the process of vector fusion processing of the initial recommendation basis vector and the reference recommendation basis vector. Any vector fusion processing process that can obtain the final recommendation basis for the current user and improve the accuracy of the recommendation of the platform's merchants can be used as the process of vector fusion processing of the initial recommendation basis vector and the reference recommendation basis vector provided by the present application. Optionally, the process of vector fusion processing includes but is not limited to: one of the following processing: vector addition processing, vector subtraction processing, vector multiplication processing, vector dot product (inner product) processing, vector cross product (outer product) processing, vector splicing processing, vector linear combination processing, and the like.

[0179] In a possible embodiment, the process of determining the initial recommendation basis for the current user based on the first historical behavior data may include: performing feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user. Based on the behavior characteristics of the current user, determining the initial recommendation basis for the current user.

[0180] Among them, the present application does not limit the process of performing feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user. Any process that can perform feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user can be used as the process of performing feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user provided by the present application. Optionally, the behavior characteristics of the current user can be obtained based on a pre-built behavior feature extraction model. Among them, the present application does not limit the behavior characteristics of the current user. Any behavior characteristics can be used as the user behavior characteristics provided by the present application. Optionally, the user behavior characteristics include, but are not limited to: at least one of frequency characteristics, time characteristics, consumption characteristics, interaction characteristics, behavior sequence characteristics, merchant characteristics and other characteristics.

[0181] Among them, this application does not limit the process of determining the initial recommendation basis for the current user based on the behavioral characteristics of the current user. Optionally, the user intention of the current user can be determined based on the behavioral characteristics of the current user, and the initial recommendation basis for the current user can be determined based on the user intention of the current user.

[0182] Figure 7 A schematic diagram of the structure of the recommendation device for merchants on the platform provided by this application, such as Figure 7 As shown, the recommendation device 70 for merchants of the platform provided in this embodiment includes:

[0183] The first determination module 701 is used to obtain first historical behavior data when the current user interacts with the platform, and determine the initial recommendation basis of the current user according to the first historical behavior data;

[0184] The second determination module 702 is used to determine the reference recommendation basis of the current user according to the first historical behavior data and the pre-constructed reference basis recommendation pool; wherein the reference basis recommendation pool is a data pool pre-constructed based on the big model and the second historical behavior data when the historical user interacts with the platform, and is used to determine the reference recommendation basis of the current user;

[0185] The recommendation module 703 is used to determine the final recommendation basis of the current user according to the initial recommendation basis and the reference recommendation basis; and determine and present at least one merchant to be recommended according to the final recommendation basis.

[0186] In a possible implementation, the device further includes a construction module for acquiring second historical behavior data of historical users interacting with the platform; based on the big model, a reference recommendation basis pool is determined according to the second historical behavior data.

[0187] In one possible implementation, a construction module is specifically used to determine a set of seed users from historical users based on the second historical behavior data; wherein the seed user set includes at least one seed user; obtain the third historical behavior data of each seed user in the seed user set when interacting with the platform, and based on the big model, determine the reference recommendation basis for each seed user based on the third historical behavior data; and determine a reference basis recommendation pool based on the reference recommendation basis for each seed user.

[0188] In one possible implementation, the construction module is further specifically used to determine the interaction behavior sequence of each seed user based on the third historical behavior data; based on the big model, determine the preference prompt word of each seed user according to the interaction behavior sequence of each seed user; based on the big model, determine the reference recommendation basis of each seed user according to the preference prompt word of each seed user.

[0189] In a possible implementation, the construction module is further specifically used to determine a representative seed user set from historical users based on the second historical behavior data; determine an important seed user set from historical users based on the second historical behavior data; and determine a seed user set based on the representative seed user set and the important seed user set.

[0190] In a possible implementation, the construction module is further specifically used to process the second historical behavior data based on a user behavior vector generation algorithm to obtain a user behavior vector for each user; based on a clustering algorithm, cluster the user behavior vector of each user to obtain at least one user cluster set; the user cluster set includes at least one user; according to a preset extraction strategy, extract at least one user from each user cluster set, determine the extracted user as a representative seed user, and determine a set consisting of representative seed users extracted from each user cluster set as a representative seed user set.

[0191] In a possible implementation, the preset extraction strategy includes: a preset ratio extraction strategy; a construction module, which is further specifically used to determine, for each user cluster set, the number of users included in the user cluster set; determine the number of users to be extracted in the user cluster set according to the extraction ratio indicated by the preset ratio extraction strategy and the number of users included in the user cluster set; and extract at least one user from the user cluster set according to the number of users to be extracted in the user cluster set.

[0192] In one possible implementation, the preset extraction strategy includes: a preset quantity extraction strategy; a construction module, which is further specifically used to determine the cluster center point of the user cluster set for each user cluster set; determine the distance between each user in the user cluster set and the cluster center point of the user cluster set; extract at least one user from the user cluster set according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set.

[0193] In a possible implementation, the construction module is further specifically used to determine the number of users included in the user cluster set; if the extraction number indicated by the preset number extraction strategy is less than the number of users included in the user cluster set, then at least one user is extracted from the user cluster set based on the distance between each user in the user cluster set and the cluster center point; wherein the number of extracted users is equal to the extraction number indicated by the preset number extraction strategy; if the extraction number indicated by the preset number extraction strategy is greater than or equal to the number of users included in the user cluster set, then all users are extracted from the user cluster set.

[0194] In a possible implementation, the construction module is further specifically used to determine the number of user interaction behaviors of each user based on the second historical behavior data; and determine a set of important seed users from historical users and current users based on a preset sampling strategy and the number of user interaction behaviors of each user.

[0195] In a possible implementation, a sampling strategy is preset to indicate a threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors; wherein the total number of user interaction behaviors is the sum of the number of user interaction behaviors of each user; a construction module is further specifically used to calculate the ratio of the number of user interaction behaviors of each user to the total number of user interaction behaviors; a user whose calculated ratio is greater than the threshold value of the ratio of the number of user interaction behaviors to the total number of user interaction behaviors is determined as an importance seed user in the importance seed user set.

[0196] In a possible implementation, the sampling strategy is preset and a threshold value of the number of user interaction behaviors is also indicated; the construction module is further specifically used to screen users whose number of user interaction behaviors is greater than the threshold value of the number of user interaction behaviors.

[0197] In a possible implementation manner, the ratio of the number of representative seed users included in the representative seed user set to the number of important seed users included in the important seed user set is a preset ratio threshold.

[0198] In one possible implementation, the pre-constructed reference basis recommendation pool includes a reference recommendation basis for each seed user and a user behavior vector for each seed user; the second determination module 702 is specifically used to process the first historical behavior data based on a user behavior vector generation algorithm to obtain a first user behavior vector for the current user; based on a vector similarity algorithm, calculate the similarity between the first user behavior vector and the user behavior vector of each seed user included in the pre-constructed reference recommendation basis pool; and based on the calculated similarity, determine the reference recommendation basis for the current user from the reference recommendation basis for each seed user included in the pre-constructed reference recommendation basis pool.

[0199] In one possible implementation, the recommendation module 703 is specifically used to process the initial recommendation basis and the reference recommendation basis based on a vector generation algorithm to obtain an initial recommendation basis vector and a reference recommendation basis vector; perform vector fusion processing on the initial recommendation basis vector and the reference recommendation basis vector to obtain the final recommendation basis for the current user.

[0200] In a possible implementation, the first determination module 701 is specifically configured to perform feature extraction processing on the first historical behavior data to obtain the behavior characteristics of the current user; and determine the initial recommendation basis for the current user according to the behavior characteristics of the current user.

[0201] The recommendation device for merchants on the platform provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be described in detail here.

[0202] Figure 8 This is a schematic diagram of the structure of the electronic device provided in this application. Figure 8 As shown, the electronic device 80 provided in this embodiment includes: at least one processor 801 and a memory 802. Optionally, the device 80 also includes a communication component 803. The processor 801, the memory 802 and the communication component 803 are connected via a bus 804.

[0203] In a specific implementation process, at least one processor 801 executes the computer-executable instructions stored in the memory 802, so that at least one processor 801 executes the above method.

[0204] The specific implementation process of the processor 801 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0205] In the above embodiments, it should be understood that the processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the invention can be directly implemented as a hardware processor, or can be implemented by a combination of hardware and software modules in the processor.

[0206] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk storage.

[0207] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.

[0208] The present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0209] The present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0210] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0211] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0212] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0213] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0214] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0215] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0216] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0217] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for recommending merchants on a platform, characterized in that: include: Acquire first historical behavior data when the current user interacts with the platform, and determine an initial recommendation basis for the current user based on the first historical behavior data; Determine the reference recommendation basis of the current user according to the first historical behavior data and the pre-constructed reference basis recommendation pool; wherein the reference basis recommendation pool is a data pool pre-constructed based on the big model and the second historical behavior data when the historical user interacts with the platform, and is used to determine the reference recommendation basis of the current user; According to the initial recommendation basis and the reference recommendation basis, a final recommendation basis for the current user is determined; and according to the final recommendation basis, at least one merchant to be recommended is determined and presented.

2. The method according to claim 1, characterized in that The construction process of the reference recommendation pool includes: Acquire second historical behavior data when the historical user interacts with the platform; Based on the big model, a reference recommendation basis pool is determined according to the second historical behavior data.

3. The method according to claim 2, characterized in that Based on the big model and the second historical behavior data, a reference recommendation basis pool is determined, including: Determine a seed user set from the historical users according to the second historical behavior data; wherein the seed user set includes at least one seed user; Acquire third historical behavior data of each seed user in the seed user set when interacting with the platform, and determine a reference recommendation basis for each seed user based on the big model and the third historical behavior data; The reference basis recommendation pool is determined according to the reference recommendation basis of each sub-user.

4. The method according to claim 3, characterized in that: Based on the big model and according to the third historical behavior data, a reference recommendation basis for each seed user is determined, including: Determining an interaction behavior sequence of each sub-user according to the third historical behavior data; Based on the large model, determining the preferred prompt words of each sub-user according to the interaction behavior sequence of each sub-user; Based on the big model, a reference recommendation basis for each seed user is determined according to the preference prompt words of each seed user.

5. The method according to claim 3, characterized in that: Determining a set of seed users from the historical users according to the second historical behavior data includes: Determining a representative seed user set from the historical users according to the second historical behavior data; Determining an important seed user set from the historical users according to the second historical behavior data; The seed user set is determined according to the representative seed user set and the importance seed user set.

6. The method according to claim 5, characterized in that Determining a representative seed user set from the historical users according to the second historical behavior data includes: Based on a user behavior vector generation algorithm, the second historical behavior data is processed to obtain a user behavior vector for each user; Based on a clustering algorithm, clustering is performed on the user behavior vector of each user to obtain at least one user cluster set; the user cluster set includes at least one user; According to a preset extraction strategy, at least one user is extracted from each user cluster set, the extracted user is determined as the representative seed user, and the set consisting of the representative seed users extracted from each user cluster set is determined as the representative seed user set.

7. The method according to claim 6, characterized in that The preset extraction strategy includes: a preset ratio extraction strategy; extracting at least one user from each user cluster set according to the preset extraction strategy, including: For each user cluster set, determining the number of users included in the user cluster set; Determining the number of users to be extracted from the user cluster set according to the extraction ratio indicated by the preset ratio extraction strategy and the number of users included in the user cluster set; At least one user is extracted from the user cluster set according to the number of users to be extracted from the user cluster set.

8. The method according to claim 6, characterized in that The preset extraction strategy includes: a preset number extraction strategy; extracting at least one user from each user cluster set according to the preset extraction strategy, including: For each user cluster set, determine the cluster center point of the user cluster set; Determine the distance between each user in the user cluster set and the cluster center point of the user cluster set; At least one user is extracted from the user cluster set according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set.

9. The method according to claim 8, characterized in that Extracting at least one user from the user cluster set according to the extraction quantity indicated by the preset quantity extraction strategy and the distance between each user in the user cluster set and the cluster center point of the user cluster set, including: Determining the number of users included in the user cluster set; If the extraction quantity indicated by the preset extraction strategy is less than the number of users included in the user cluster set, extract at least one user from the user cluster set according to the distance between each user in the user cluster set and the cluster center point; wherein the number of users extracted is equal to the extraction quantity indicated by the preset extraction strategy; If the extraction quantity indicated by the preset quantity extraction strategy is greater than or equal to the number of users included in the user cluster set, all users are extracted from the user cluster set.

10. The method according to claim 5, characterized in that Determining an important seed user set from the historical users according to the second historical behavior data includes: determining the number of user interaction behaviors of each user according to the second historical behavior data; The important seed user set is determined from the historical users according to a preset sampling strategy and the number of user interaction behaviors of each user.

11. The method according to claim 10, characterized in that The preset sampling strategy indicates a threshold value of the ratio of the number of user interaction behaviors to the number of all user interaction behaviors; wherein the number of all user interaction behaviors is the sum of the number of user interaction behaviors of each user; According to a preset sampling strategy and the number of user interaction behaviors of each user, determining the important seed user set from the historical users includes: Calculate the ratio of the number of user interaction behaviors of each user to the total number of user interaction behaviors; The users whose calculated ratio is greater than a threshold of the ratio of the number of user interaction behaviors to the number of all user interaction behaviors are determined as importance seed users in the importance seed user set.

12. The method according to claim 11, characterized in that The preset sampling strategy also indicates a threshold value of the number of user interaction behaviors; Before calculating the proportion of the number of user interaction behaviors of each user to the total number of user behaviors, the method further includes: Screen users whose number of user interaction behaviors is greater than the user interaction behavior number threshold.

13. The method according to claim 5, characterized in that A ratio of the number of the representative seed users included in the representative seed user set to the number of the important seed users included in the important seed user set is a preset ratio threshold.

14. The method according to any one of claims 1 to 13, characterized in that: The pre-constructed reference basis recommendation pool includes a reference recommendation basis for each seed user and a user behavior vector for each seed user; Determining the reference recommendation basis of the current user according to the first historical behavior data and the pre-built reference basis recommendation pool includes: Based on a user behavior vector generation algorithm, the first historical behavior data is processed to obtain a first user behavior vector of the current user; Based on a vector similarity algorithm, calculating the similarity between the first user behavior vector and the user behavior vector of each of the seed users included in the pre-built reference recommendation basis pool; According to the calculated similarity, the reference recommendation basis of the current user is determined from the reference recommendation basis of each sub-user included in the pre-constructed reference recommendation basis pool.

15. The method according to any one of claims 1 to 13, characterized in that: Determining the final recommendation basis for the current user based on the initial recommendation basis and the reference recommendation basis includes: Based on a vector generation algorithm, the initial recommendation basis and the reference recommendation basis are processed to obtain an initial recommendation basis vector and a reference recommendation basis vector; Vector fusion processing is performed on the initial recommendation basis vector and the reference recommendation basis vector to obtain a final recommendation basis for the current user.

16. The method according to any one of claims 1 to 13, characterized in that: Determine the initial recommendation basis for the current user based on the first historical behavior data, including: Performing feature extraction processing on the first historical behavior data to obtain the behavior features of the current user; An initial recommendation basis for the current user is determined according to the behavior characteristics of the current user.

17. A recommendation device for merchants on a platform, characterized in that: include: A first determination module, used to obtain first historical behavior data when the current user interacts with the platform, and determine the initial recommendation basis for the current user based on the first historical behavior data; A second determination module, configured to obtain second historical behavior data of a historical user when interacting with the platform, and determine a reference recommendation basis for the current user based on the second historical behavior data and the first historical behavior data; A recommendation module, used to determine a final recommendation basis for the current user based on the initial recommendation basis and the reference recommendation basis; And according to the final recommendation basis, at least one merchant to be recommended is determined and presented.

18. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 16.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method according to any one of claims 1 to 16.

20. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 16 when being executed by a processor.

Images