CN112100511A

CN112100511A - Preference degree data obtaining method and device and electronic equipment

Info

Publication number: CN112100511A
Application number: CN202011305501.2A
Authority: CN
Inventors: 衣建中
Original assignee: Zhejiang Koubei Network Technology Co Ltd
Current assignee: Zhejiang Koubei Network Technology Co Ltd
Priority date: 2020-11-18
Filing date: 2020-11-18
Publication date: 2020-12-18
Anticipated expiration: 2040-11-18
Also published as: CN112100511B

Abstract

The embodiment of the application provides a method for obtaining preference degree data, which comprises the following steps: obtaining positive sample data of a scene strategy; constructing scene strategy negative sample data; according to the scene strategy positive sample data and the scene strategy negative sample data, obtaining a first target crowd label vector corresponding to the first crowd label data, a first target scene label vector corresponding to the first scene label data and a first target object label vector corresponding to the first object label data; according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector, preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data is obtained. The preference degree data obtaining method provided by the embodiment of the application solves the problem of how to obtain the preference degree data of the user for the object.

Description

Preference degree data obtaining method and device and electronic equipment

Technical Field

The application relates to the technical field of computers, in particular to a preference degree data obtaining method. The application also relates to a preference degree data obtaining device, an electronic device and a storage medium. In addition, the application also relates to an object recommendation method, an object recommendation device, electronic equipment and a storage medium.

Background

With the rapid development of internet technology and mobile payment technology, the online service platform also starts to develop at a high speed. The online service platform often attracts a user to consume to a corresponding object of the user when the user accesses an APP (Application) corresponding to the online service platform, for example, the takeaway service platform recommends a corresponding takeaway object to the user on a takeaway APP top page when the user accesses the takeaway APP. In order to be able to attract users to consume, objects that meet the user preferences need to be recommended to the users. At this time, in order to recommend the object meeting the user preference to the user, it is necessary to first obtain the preference degree data of the user for the object, and then recommend the object meeting the user preference to the user according to the preference degree data of the user for the object.

Therefore, how to obtain the preference degree data of the user for the object becomes a serious issue in recommending the object according with the preference of the user to the user.

Disclosure of Invention

The application provides a preference degree data obtaining method, a preference degree data obtaining device, electronic equipment and a storage medium, so as to obtain preference degree data of a user on an object.

The embodiment of the application provides a method for obtaining preference degree data, which comprises the following steps: obtaining scene strategy positive sample data, wherein the scene strategy positive sample data comprises first crowd label data, first scene label data and first object label data with historical incidence relation; constructing scene strategy negative sample data opposite to the scene strategy positive sample data, wherein the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation; according to the scene strategy positive sample data and the scene strategy negative sample data, obtaining a first target crowd label vector corresponding to the first crowd label data, a first target scene label vector corresponding to the first scene label data and a first target object label vector corresponding to the first object label data; according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector, preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data is obtained.

Optionally, the obtaining of positive sample data of the scene policy includes: obtaining a scene strategy sample with the quantity ratio exceeding a preset ratio threshold value in the full-scale scene strategy sample data, and determining the scene strategy sample with the quantity ratio exceeding the preset ratio threshold value as the scene strategy positive sample data.

Optionally, the method further includes: obtaining historical behavior data of a user; extracting user data, scene data and object data from the user historical behavior data; obtaining crowd label data corresponding to the user data; obtaining scene tag data corresponding to the scene data; obtaining object tag data corresponding to the object data; generating scene strategy sample data according to the crowd label data, the scene label data and the object label data; and determining all generated scene strategy sample data as the full-scale scene strategy sample data.

Optionally, the method further includes: obtaining a plurality of crowd tag data; obtaining a plurality of scene tag data; obtaining a plurality of object tag data; selecting one crowd tag data from the plurality of crowd tag data, selecting one scene tag data from the plurality of scene tag data, and selecting one object tag data from the plurality of object tag data; generating scene strategy sample data according to randomly selected crowd label data, randomly selected scene label data and randomly selected object label data; determining all generated scene strategy sample data as the full-scale scene strategy sample data; and the scene strategy sample with the number ratio exceeding a preset ratio threshold is a scene strategy sample generated according to the historical behavior data of the user.

Optionally, the constructing of the scene policy negative sample data opposite to the scene policy positive sample data includes: replacing the first crowd tag data included in the scene policy positive sample data with the second crowd tag data different from the first crowd tag data to obtain the scene policy negative sample data; or, the first crowd tag data is the same as the second crowd tag data, the first scene tag data is the same as the second scene tag data, and the constructing of scene policy negative sample data relative to the scene policy positive sample data includes: replacing the first object tag data included in the scene policy positive sample data with the second object tag data different from the first object tag data to obtain the scene policy negative sample data; and the negative sample data of any scene strategy is different from the positive sample data of any scene strategy.

Optionally, the obtaining, according to the scene policy positive sample data and the scene policy negative sample data, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data includes: obtaining a first initial crowd label vector corresponding to the first crowd label data, obtaining a first initial scene label vector corresponding to the first scene label data, and obtaining a first initial object label vector corresponding to the first object label data; obtaining a second initial crowd label vector corresponding to the second crowd label data, obtaining a second initial scene label vector corresponding to the second scene label data, and obtaining a second initial object label vector corresponding to the second object label data; and obtaining the first target crowd label vector, the first target scene label vector and the first target object label vector according to the first initial crowd label vector, the first initial scene label vector, the first initial object label vector, the second initial crowd label vector, the second initial scene label vector and the second initial object label vector.

Optionally, the obtaining a first initial crowd tag vector corresponding to the first crowd tag data includes: obtaining a first initial population sub-label vector corresponding to each first population sub-label data included in the first population label data, and performing weighted average processing on all the first initial population sub-label vectors to obtain the first initial population label vectors; or, the obtaining a first initial scene tag vector corresponding to the first scene tag data includes: obtaining a first initial scene sub-label vector corresponding to each first scene sub-label data included in the first scene label data, and performing weighted average processing on all the first initial scene sub-label vectors to obtain the first initial scene label vectors; or, the obtaining a first initial object tag vector corresponding to the first object tag data includes: obtaining a first initial object sub-label vector corresponding to each first object sub-label data included in the first object label data, and performing weighted average processing on all the first initial object sub-label vectors to obtain the first initial object label vectors; or, the obtaining a second initial crowd tag vector corresponding to the second crowd tag data includes: obtaining a second initial population sub-label vector corresponding to each second population sub-label data included in the second population label data, and performing weighted average processing on all second initial population sub-label vectors to obtain a second initial population label vector; or, the obtaining a second initial scene tag vector corresponding to the second scene tag data includes: obtaining a second initial scene sub-label vector corresponding to each second scene sub-label data included in the second scene label data, and performing weighted average processing on all the second initial scene sub-label vectors to obtain a second initial scene label vector; or, the obtaining a second initial object tag vector corresponding to the second object tag data includes: and obtaining a second initial object sub-label vector corresponding to each second object sub-label data included in the second object label data, and performing weighted average processing on all the second initial object sub-label vectors to obtain the second initial object label vectors.

Optionally, the number of at least one of the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector is plural; the obtaining the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector according to the first initial crowd tag vector, the first initial scene tag vector, the first initial object tag vector, the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector includes: taking the first initial population label vector, the first initial scene label vector, the first initial object label vector, the second initial population label vector, the second initial scene label vector and the second initial object label vector as input parameters of a loss function, and obtaining the first initial population label vector, the first initial scene label vector and the first initial object label vector when a loss function value is a minimum value; and determining a first initial crowd tag vector, a first initial scene tag vector and a first initial object tag vector when the loss function value is a minimum value as the first target crowd tag vector, the first target scene tag vector and the first target object tag vector.

Optionally, the obtaining the first initial population label vector, the first initial scene label vector, the first initial object label vector, the second initial population label vector, the second initial scene label vector, and the second initial object label vector when the loss function value is the minimum value, as an input parameter of the loss function, includes: calculating a first modulo result of a first distance value between the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector; calculating a second modulo result of a second distance value between the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector; calculating a modulus result difference between the first modulus result and the second modulus result; calculating a third sum between the modulus-taking result difference and the hinge function value; determining a maximum value of the third sum value compared to a preset loss function value as a loss function value; determining the minimum value of all the obtained loss function values as the minimum value of the loss function; and obtaining a first initial crowd label vector, a first initial scene label vector and a first initial object label vector corresponding to the minimum value of the loss function.

Optionally, the calculating a first modulo result of a first distance value between the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector includes: calculating a first sum between the first initial crowd tag vector and the first initial scene tag vector; calculating a first difference between the first sum and the first initial object tag vector; a first modulo result for the first difference is obtained.

Optionally, the calculating a second modulo result of a second distance value between the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector includes: calculating a second sum between the second initial crowd tag vector and the second initial scene tag vector; calculating a second difference between the second sum and the second initial object tag vector; obtaining a second modulo result for the second difference.

Optionally, the obtaining, according to the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector, preference degree data of a crowd corresponding to the first crowd tag data for the first object tag data in a first scene corresponding to the first scene tag data includes: calculating a third modulo result of a third distance value between the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector; and obtaining preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data according to the third modulus result.

Optionally, the first scene tag data includes first temporal sub-tag data, first spatial sub-tag data, and first environmental sub-tag data; or, the second scene tag data includes second temporal sub-tag data, second spatial sub-tag data, and second environmental sub-tag data.

In another aspect of the embodiments of the present application, there is also provided a preference degree data obtaining apparatus, including: the system comprises a positive sample data obtaining unit, a positive sample data obtaining unit and a negative sample data obtaining unit, wherein the positive sample data obtaining unit is used for obtaining scene strategy positive sample data which comprises first crowd tag data, first scene tag data and first object tag data with historical incidence relation; the negative sample data construction unit is used for constructing scene strategy negative sample data relative to the scene strategy positive sample data, and the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have a historical incidence relation; a tag vector obtaining unit, configured to obtain, according to the scene policy positive sample data and the scene policy negative sample data, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data; a preference degree obtaining unit, configured to obtain, according to the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector, preference degree data of a crowd corresponding to the first crowd tag data for the first object tag data in a first scene corresponding to the first scene tag data.

In another aspect of the embodiments of the present application, there is also provided an object recommendation method, including: obtaining target user information; obtaining target crowd label data corresponding to the target user information; obtaining target scene label data; recommending target object data corresponding to target object tag data to a user corresponding to the target user information according to pre-obtained preference degree data of a crowd corresponding to the first target crowd tag data in a first target scene corresponding to the first target scene tag data; wherein preference degree data of a crowd corresponding to the target first-person group tag data for first candidate object tag data in a target first scene corresponding to first target scene tag data is obtained according to a first target first-person group tag vector corresponding to the target first-person group tag data, a first target scene tag vector corresponding to the target first-person group tag data, and a first target candidate object tag vector corresponding to the first candidate object tag data, the first target crowd tag vector, the first target scene tag vector, and the first target candidate object tag vector being obtained according to scene policy positive sample data and scene policy negative sample data, the scene policy positive sample data including the first target crowd tag data, the target first scene tag data, and the first candidate object tag data having a historical association relationship, the scene policy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation, and the scene policy negative sample data is negative sample data opposite to the scene policy positive sample data.

Optionally, the recommending, according to preference degree data of a crowd corresponding to the target crowd tag data obtained in advance for candidate object tag data in a target scene corresponding to the target scene tag data, target object data corresponding to the target object tag data to a user corresponding to the target user information includes: obtaining candidate object tag data corresponding to the highest value corresponding to the preference degree data; determining candidate object tag data corresponding to the highest value as the target object tag data; obtaining the target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

Optionally, the recommending, according to preference degree data of a crowd corresponding to the target crowd tag data obtained in advance for candidate object tag data in a target scene corresponding to the target scene tag data, target object data corresponding to the target object tag data to a user corresponding to the target user information includes: obtaining candidate object tag data corresponding to preference degree data exceeding a preset value; determining candidate object tag data corresponding to preference degree data exceeding a preset value as the target object tag data; obtaining the target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

Optionally, the obtaining target user information includes: obtaining the target user information provided by a client corresponding to the user; the recommending target object data corresponding to the target object tag data to the user corresponding to the target user information includes: and sending the target object data to a client corresponding to the user.

Optionally, the obtaining target user information includes: responding to a trigger operation aiming at a target application, and obtaining target user information for logging in the target application; the recommending target object data corresponding to the target object tag data to the user corresponding to the target user information includes: and displaying the target object data on the page of the target application.

Optionally, the obtaining target crowd tag data corresponding to the target user information includes: sending a first request message for requesting to obtain target crowd label data corresponding to the target user information to a server; and obtaining the target crowd label data provided by the server.

Optionally, the obtaining target scene tag data includes: obtaining at least one of target time data, target space data and target environment data through a running device running the target application; obtaining at least one of the following seed label data: target time sub-label data corresponding to the target time data, target space sub-label data corresponding to the target space data, and target environment sub-label data corresponding to the target environment data; and generating the target scene label data according to the target time sub-label data, the target space sub-label data and the target environment sub-label data.

Optionally, the method further includes: sending a second request message for requesting to obtain preference degree data of the crowd corresponding to the target crowd tag data to the candidate object tag data in the target scene corresponding to the target scene tag data to a server; and obtaining preference degree data of the crowd corresponding to the target crowd label data provided by the server side for candidate object label data in the target scene corresponding to the target scene label data.

In another aspect of the embodiments of the present application, there is also provided an object recommendation apparatus, including: a target user information obtaining unit for obtaining target user information; the target crowd tag data obtaining unit is used for obtaining target crowd tag data corresponding to the target user information; a target scene tag data obtaining unit for obtaining target scene tag data; the target object data recommending unit is used for recommending target object data corresponding to the target object tag data to a user corresponding to the target user information according to pre-obtained preference degree data of a crowd corresponding to the first target crowd tag data to the first candidate object tag data in a first target scene corresponding to the first target scene tag data; wherein preference degree data of a crowd corresponding to the target first-person group tag data for first candidate object tag data in a target first scene corresponding to first target scene tag data is obtained according to a first target first-person group tag vector corresponding to the target first-person group tag data, a first target scene tag vector corresponding to the target first-person group tag data, and a first target candidate object tag vector corresponding to the first candidate object tag data, the first target crowd tag vector, the first target scene tag vector, and the first target candidate object tag vector being obtained according to scene policy positive sample data and scene policy negative sample data, the scene policy positive sample data including the first target crowd tag data, the target first scene tag data, and the first candidate object tag data having a historical association relationship, the scene policy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation, and the scene policy negative sample data is negative sample data opposite to the scene policy positive sample data.

In another aspect of the embodiments of the present application, an electronic device is further provided, which includes: a processor; and the memory is used for storing the program of the method, and the device executes the method provided by the embodiment after being powered on and running the program of the method through the processor.

In another aspect of the embodiments of the present application, a storage medium is further provided, where the storage medium stores a computer program, and the computer program is executed by a processor to perform the method provided in the embodiments of the present application.

Compared with the prior art, the method has the following advantages:

the preference degree data obtaining method provided by the embodiment of the application comprises the steps of firstly, obtaining scene strategy positive sample data, wherein the scene strategy positive sample data comprises first crowd label data, first scene label data and first object label data with historical incidence relation; secondly, constructing scene strategy negative sample data relative to the scene strategy positive sample data, wherein the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation; and thirdly, according to the scene strategy positive sample data and the scene strategy negative sample data, obtaining a first target crowd label vector corresponding to the first crowd label data, a first target scene label vector corresponding to the first scene label data and a first target object label vector corresponding to the first object label data. And finally, according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector, obtaining preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data. According to the preference degree data obtaining method provided by the embodiment of the application, preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data can be obtained, and therefore the problem of how to obtain the preference degree data of the user to the object is solved. In addition, the object preferred by the user can be determined according to the preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data.

Drawings

Fig. 1 is a scene schematic diagram of a preference degree data obtaining method provided by the present application.

Fig. 2 is a flowchart of a preference degree data obtaining method provided in the first embodiment of the present application.

Fig. 3 is a schematic diagram of a tag vector obtaining method provided in the first embodiment of the present application.

Fig. 4 is a table diagram of preference degrees provided in the first embodiment of the present application.

Fig. 5 is a schematic diagram of a preference degree data obtaining apparatus provided in a second embodiment of the present application.

Fig. 6 is a flowchart of an object recommendation method provided in a third embodiment of the present application.

Fig. 7 is a schematic diagram of an object recommendation method provided in a fourth embodiment of the present application.

Fig. 8 is a schematic diagram of an electronic device provided in an embodiment of the present application.

Detailed Description

In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. This application is capable of implementation in many different ways than those herein set forth and of similar import by those skilled in the art without departing from the spirit of this application and is therefore not limited to the specific implementations disclosed below.

In order to more clearly show the method for obtaining preference degree data provided by the embodiment of the present application, an application scenario of the method for obtaining preference degree data provided by the embodiment of the present application is first introduced.

Please refer to fig. 1, which is a scene diagram illustrating a method for obtaining preference level data according to the present application.

In practical applications, the preference degree data for the designated object tag data in the designated scene corresponding to the designated scene tag data for obtaining the crowd corresponding to the designated crowd tag data may be stored in a database for storing data in the server 101. The Database may be a Database such as HBase (HBase-Hadoop Database, distributed, column-oriented open source Database).

The crowd tag data, the scene tag data, and the object tag data are preset tag data. The crowd tag data is tag data for identifying attributes of the crowd, the attributes of the crowd identified by the crowd tag data include, but are not limited to, gender, age, work, marital status, academic history, etc. the attributes of the crowd identified by the crowd tag data include at least one of gender, age, work, marital status, academic history, etc. of the crowd, and the attributes of the crowd identified by different crowd tag data are different. The gender, age, job, marital status, academic history, etc. of the population within the population attributes identified by the population label data may be further divided according to daily habits or experiences, such as: the age of the population in the population attributes identified by the population label data may be further divided into children, teenagers, middle-aged people, elderly people, etc., and may be further divided into different age groups: 0-8 years old, 9-18 years old, 19-40 years old, 41-65 years old, and the like. The work of the crowd in the crowd attributes identified by the crowd label data can be further divided into white collar, blue collar, gold collar and the like, and can also be further divided into different work types, such as: doctors, teachers, lawyers, attendants, etc. The crowd tag data may be: middle aged, married, and aged; the method can also comprise the following steps: young, not married, students, men; the following steps can be also included: 20-25 years old, not married, teacher, woman, this family. The scene tag data is tag data for identifying scene attributes, the scene attributes identified by the scene tag data include, but are not limited to, time, space, environment, and the like to which the scene belongs, the scene attributes identified by the scene tag data include at least one of time, space, environment, and the like to which the scene belongs, and the scene attributes identified by different scene tag data are different. Wherein, the space is generally a corresponding position of the space, such as: shanghai, Beijing Haihu district, and Beijing Kogyo-rising district, No. 5 building 3 unit. The time, space, environment, etc. to which the scene belongs in the scene attribute identified by the scene tag data may be further divided according to daily habits or experiences, such as: the time of the scene in the scene attribute may be further divided into morning, noon, afternoon, evening, and the like, or may be further divided into time periods: 0-6 hours, 6-12 hours, 12-18 …. The scene tag data may be: noon, last sea, lunch time; the method can also comprise the following steps: noon, beijing, evening, sunny day; the following steps can be also included: 12-18 hours, Chengdu, at home, and in rain.

The object tag data is tag data for identifying attributes of the object tag data, the attributes of the object identified by the object tag data include, but are not limited to, a category, a rank, a price, and the like of the object, the attributes of the object identified by the object tag data include at least one of the category, the rank, the price, and the like of the object, and the attributes of the object identified by different object tag data are different. The object type, ranking, price, and the like in the object attribute identified by the object tag data may be further classified according to daily habits or experiences, and taking the object as a commodity as an example, the object type in the object attribute identified by the object tag data may be further classified into food, fruit, toys, meat, and the like. If the object is a food, the object type in the object attribute identified by the object tag data can be further divided into fast food, Chinese fast food, wheaten food, bread dessert, milky tea, fruit juice, and the like. Taking the object as a business point as an example, the object type in the object attribute identified by the object tag data can be further divided into fast food restaurants, clothing stores, jewelry stores and the like. The object tag data may be: snack, Shanghai region No. 2, everyone 50 Yuan; the method can also comprise the following steps: the 3 rd name of the wheaten food and the Haihu district, 68 Yuan everywhere.

The preference degree data for the designated object tag data is data indicating a preference degree for the designated object tag data, and specifically may be a selection probability for the designated object tag data. The server 101 is generally a server of an online service platform, and the server is generally a server or a server cluster in a specific implementation manner. The online service platform includes, but is not limited to, a take-out service platform, a power provider service platform, a short video service platform, and the like.

When a target user accesses the client 102, a target object acquisition request message for requesting to acquire target object data is sent to the server 102. The server 101 may first obtain a target user account of a target user logging in the client 101, and further obtain target crowd tag data corresponding to the target user according to the user account of the target user. That is, first, in response to a trigger operation for a target application, target user information for logging in the target application is obtained. Then, the server 101 obtains the target user account provided by the client 102. The client 102 is generally a client of the online service platform, and specifically, an application or software installed on the computing device. The user account is generally a user ID (Identity Document). The specific implementation manners of obtaining the target crowd label data corresponding to the target user according to the user account of the target user include at least the following two ways:

firstly, identity user characteristic information stored by a target user is obtained according to a user account of the target user, wherein the identity user characteristic information includes but is not limited to sex information, age information, work information, marital status information, academic history information and the like. And then, according to the identity user characteristic information stored by the target user, performing label data matching with preset crowd label data to obtain target crowd label data corresponding to the target user. That is, according to the identity user feature information, gender, age, work, marital status, academic history, and the like of the crowd in the crowd attributes identified by the crowd tag data are matched, and the target crowd tag data corresponding to the target user is obtained. And secondly, acquiring the crowd label data selected by the target user as the target crowd label data corresponding to the target user according to the user account of the target user.

In addition, when the target user accesses the client 102, the server 101 obtains access time information and access location information of the target user accessing the client 102 through the access log stored in the client 102, and determines corresponding environment information under the current access time information and the access location information according to the access time information and the access location information. The visiting location information may be specific longitude and latitude information, or may be specific geographic location information, such as: and a certain district in a certain city has a No. 5 building unit. It is possible to determine whether the location where the user is located under the current visit time information is home or company from the visit location information. According to the access time information and the access position information, the corresponding environment information under the current access time information and the access position information can be further determined, such as: after the current access address is determined to be a certain area in a certain city according to the direction address information and the current access time is determined to be 12 hours on a certain month and a certain day of a certain year according to the access time information, the weather of the certain area in the certain city on the certain day 12 hours on the certain month and the certain day of the certain year can be further determined, namely, the current access time information and the corresponding environment information under the access position information are obtained.

After the server 101 obtains the access time information, the access location information, and the access time information and the environment information corresponding to the access location information, tag data matching may be performed with preset scene tag data according to the access time information, the access location information, and the environment information corresponding to the access time information and the access location information, so as to obtain the scene tag data corresponding to the target user. That is, the time, space, environment, and the like to which the scene belongs in the scene attribute identified by the scene tag data are matched according to the access time information, the access location information, and the access time information and the environment information corresponding to the access location information, so as to obtain the scene tag data corresponding to the target user.

After obtaining the target population tag data corresponding to the target user and the scene tag data corresponding to the target user, the server 101 may match preference degree data of the population corresponding to the specified population tag data to the specified object tag data in the specified scene corresponding to the specified scene tag data according to the target population tag data corresponding to the target user and the scene tag data corresponding to the target user, so as to obtain preference degree data of the population corresponding to the target population tag data to the target object tag data in the target scene corresponding to the target scene tag data. The server 101 obtains one or more target object tag data with higher preference degree data according to the preference degree data of the target object tag data of the crowd corresponding to the target crowd tag data in the target scene corresponding to the target scene tag data, obtains the target object data corresponding to the one or more target object tag data, and provides the target object data to the client 102. The client 102 displays target object data corresponding to one or more target object tag data, obtains a score recommended by a target user for the current object and feeds the score back to the server 101, and after obtaining the score recommended by the target user for the current object, the server 101 adjusts preference degree data of a crowd corresponding to the specified crowd tag data for the specified object tag data under a specified scene corresponding to the specified scene tag data according to the score.

In the embodiment of the present application, an application scenario of the preference degree data obtaining method provided in the embodiment of the present application is not specifically limited, for example: the preference degree data obtaining method provided by the application can also be executed based on the client side independently, and is not described in detail herein. The embodiment corresponding to the application scenario of the preference degree data obtaining method is provided for facilitating understanding of the preference degree data obtaining method provided by the present application, and is not used for limiting the preference degree data obtaining method provided by the present application.

First embodiment

A method for obtaining preference level data is provided in the first embodiment of the present application, and is described below with reference to fig. 2 to 4.

Fig. 2 is a flowchart of a preference degree data obtaining method provided in the first embodiment of the present application. The preference degree data obtaining method shown in fig. 2 includes: step S201 to step S204.

In step S201, scene policy positive sample data is obtained, which includes first crowd tag data, first scene tag data, and first object tag data having a historical association relationship.

In the first embodiment of the present application, the scene policy positive sample data is a scene policy sample whose number ratio exceeds a preset ratio threshold in the full-scale scene policy sample data. Correspondingly, the process of obtaining positive sample data of the scene strategy comprises the following steps: obtaining scene strategy samples with the quantity ratio exceeding a preset ratio threshold in the full-scale scene strategy sample data, and determining the scene strategy samples with the quantity ratio exceeding the preset ratio threshold as the scene strategy positive sample data.

The preset duty ratio threshold is a duty ratio threshold preset according to a priori experience, such as 5%. The scene strategy sample with the number ratio exceeding the preset ratio threshold is as follows: if the number of the full-volume scene policy sample data is 1000, the number of the full-volume scene policy sample data exceeds 1000 × 5% =50 scene policy sample data. The scene policy is a policy for determining preference degree data of a crowd corresponding to the specified crowd tag data with respect to the specified object tag data in a specified scene corresponding to the specified scene tag data, the scene is a scene where the user is currently located, and positive sample data of the scene policy is sample data for determining the scene policy.

In the first embodiment of the present application, in addition to the scene policy positive sample data, the scene policy negative sample data may also be involved, and both the scene policy positive sample data and the scene policy negative sample data may respectively include crowd tag data, scene tag data, and object tag data. For convenience of distinction, in the first embodiment of the present application, crowd tag data, scene tag data, and object tag data included in scene policy positive sample data are referred to as first crowd tag data, first scene tag data, and first object tag data, and crowd tag data, scene tag data, and object tag data included in scene policy negative sample data are referred to as second crowd tag data, second scene tag data, and second object tag data.

The crowd tag data, the scene tag data, and the object tag data are preset tag data. The crowd tag data is tag data for identifying attributes of crowd tag data, the attributes of the crowd identified by the crowd tag data include, but are not limited to, gender, age, work, marital status, academic history, etc. the attributes of the crowd identified by the crowd tag data include at least one of gender, age, work, marital status, academic history, etc. of the crowd, and the attributes of the crowd identified by different crowd tag data are different. The scene tag data is tag data for identifying attributes of the scene tag data, the attributes of the scene identified by the scene tag data include, but are not limited to, time, space, environment, and the like to which the scene belongs, the attributes of the scene identified by the scene tag data include at least one of time, space, environment, and the like to which the scene belongs, and the attributes of the scenes identified by different scene tag data are different. The object tag data is tag data for identifying attributes of the object tag data, the attributes of the object identified by the object tag data include, but are not limited to, a category, a rank, a price, and the like of the object, the attributes of the object identified by the object tag data include at least one of the category, the rank, the price, and the like of the object, and the attributes of the object identified by different object tag data are different. The crowd attribute identified by the crowd tag data, the scene attribute identified by the scene tag data and the object attribute identified by the object tag data are sub tag data of the crowd tag data, sub tag data of the scene tag data and word tag data of the object tag data respectively. In a first embodiment of the present application, the first scene tag data includes first temporal sub-tag data, first spatial sub-tag data, and first environmental sub-tag data, and the second scene tag data includes second temporal sub-tag data, second spatial sub-tag data, and second environmental sub-tag data. The preference degree data for the designated object tag data is data indicating a preference degree for the designated object tag data, and specifically may be a selection probability for the designated object tag data.

Since the scene policy positive sample data is the scene policy sample whose number ratio exceeds the preset ratio threshold in the full amount of scene policy sample data, in order to obtain the scene policy positive sample data, the following steps need to be executed in advance: first, user historical behavior data is obtained. Secondly, extracting user data, scene data and object data from the user historical behavior data. Thirdly, acquiring crowd label data corresponding to the user data; obtaining scene tag data corresponding to the scene data; object tag data corresponding to the object data is obtained. And then generating scene strategy sample data according to the crowd label data, the scene label data and the object label data. And finally, determining all the generated scene strategy sample data as full-scale scene strategy sample data. The user historical behavior data includes, but is not limited to, user consumption historical behavior data, user browsing historical behavior data, user clicking historical behavior data, and the like. The specific implementation manner of obtaining the multiple crowd label data is as follows: and obtaining user historical behavior data such as user consumption historical behavior data, user browsing historical behavior data and user clicking historical behavior data from data which correspond to one or more online service platforms and are used for storing the user data. The online service platforms include a takeout service platform, an e-commerce platform and the like.

The implementation manner of extracting user data from the user historical behavior data and obtaining crowd label data corresponding to the user data is generally as follows: data related to the user portrait is extracted from the user historical behavior data, and identity user characteristic information stored by the user, including but not limited to sex information, age information, work information, marital status information, academic history information and the like, is obtained according to the user portrait. And matching the label data with preset crowd label data according to the identity user characteristic information stored by the user to obtain the crowd label data corresponding to the user. That is, the gender, age, work, marital status, academic history, etc. of the population in the population attributes identified by the population tag data are matched according to the identity user characteristic information, and the population tag data corresponding to the user is obtained.

The implementation manner of extracting scene data from the user historical behavior data and obtaining scene tag data corresponding to the scene data is generally as follows: and extracting relevant information such as time information, position information, environment information and the like corresponding to the scene where the user is located from the historical behavior data of the user. And performing label data matching with preset scene label data according to the time information, the position information, the environment information and other related information corresponding to the scene where the user is located, so as to obtain the scene label data corresponding to the user. That is, according to the time information, the position information, the environment information and other related information corresponding to the scene where the user is located, the time, the space, the environment and the like to which the scene belongs in the scene attribute identified by the scene tag data are matched, so as to obtain the scene tag data corresponding to the user. The implementation manner of extracting object data from the user historical behavior data and obtaining object tag data corresponding to the object data is generally as follows: relevant information such as category information, ranking information and price information related to the object is extracted from the user historical behavior data. And matching the label data with preset object label data according to the related information such as the category information, the ranking information, the price information and the like of the object to obtain the object label data corresponding to the user. That is, the category, rank, price, and the like of the object in the object attribute identified by the object tag data are matched with the related information such as the category information, rank information, and price information of the object, and the object tag data corresponding to the user is obtained.

In addition, in order to obtain positive sample data of the scene policy, the following steps may also be performed in advance: a plurality of crowd tag data, a plurality of scene tag data, and a plurality of object tag data are obtained. Next, one crowd tag data is arbitrarily selected from the plurality of crowd tag data, one scene tag data is arbitrarily selected from the plurality of scene tag data, and one object tag data is arbitrarily selected from the plurality of object tag data. And thirdly, generating scene strategy sample data according to the randomly selected crowd label data, the randomly selected scene label data and the randomly selected object label data, and finally determining all the generated scene strategy sample data as full-scale scene strategy sample data. The scene strategy sample with the number ratio exceeding the preset ratio threshold is the scene strategy sample generated according to the historical behavior data of the user.

In step S202, scene policy negative sample data is constructed, which is opposite to the scene policy positive sample data, and includes second crowd tag data, second scene tag data, and second object tag data that do not have a historical association relationship.

In the first embodiment of the present application, two ways of constructing the scene policy negative sample data relative to the scene policy positive sample data are as follows:

the first mode is as follows: and replacing the first crowd label data included in the scene strategy positive sample data with second crowd label data different from the first crowd label data to obtain scene strategy negative sample data. That is, the first scene tag data is identical to the second scene tag data, and the first object tag data is identical to the second object tag data. If the first crowd label data, the first scene label data and the first object label data included in the positive sample data of the scene strategy are sequentially: first-person group tag data: male, middle year, white collar, married, first scene tag data: purdong, weekday, lunch time, first object tag data: chinese fast food, the second name of Pudong district Chinese fast food. The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data may be in sequence: second population tag data: middle year, white collar, married, second scene tag data: purdong, weekday, lunch time, second object tag data: chinese fast food, the second name of Pudong district Chinese fast food. The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data may also be in sequence: second population tag data: female, young, liberty worker, married, second scene tag data: purdong, weekday, lunch time, second object tag data: chinese fast food, the second name of Pudong district Chinese fast food. The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data may also be in sequence: second population tag data: female, young, not married, second scene tag data: purdong, weekday, lunch time, second object tag data: chinese fast food, the second name of Pudong district Chinese fast food.

At least one scene policy negative sample data can be constructed from one scene policy positive sample data, and generally, five scene policy negative sample data need to be constructed from one scene policy positive sample data. In different crowd label data, not only the crowd attributes identified by the crowd label data may be different, but also the number of the crowd attributes identified by the crowd label data may be different.

The second way is: and replacing the first object label data included in the scene strategy positive sample data with second object label data different from the first object label data to obtain scene strategy negative sample data. That is, the first crowd tag data is the same as the second crowd tag data, and the first scene tag data is the same as the second scene tag data. If the first crowd label data, the first scene label data and the first object label data included in the positive sample data of the scene strategy are sequentially: first-person group tag data: male, middle year, white collar, married, first scene tag data: purdong, weekday, lunch time, first object tag data: chinese fast food, the second name of Pudong district Chinese fast food. The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data are sequentially as follows: second population tag data: male, middle year, white collar, married, second scene tag data: purdong, weekday, lunch time, second object tag data: chinese snacks.

The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data may also be in sequence: second population tag data: male, middle year, white collar, married, second scene tag data: purdong, weekday, lunch time, second object tag data: wheaten food. The second crowd label data, the second scene label data and the second object label data included in the scene policy negative sample data may also be in sequence: second population tag data: male, middle year, white collar, married, second scene tag data: purdong, weekday, lunch time, second object tag data: milk tea and fruit juice. At least one scene policy negative sample data can be constructed from one scene policy positive sample data, and generally, five scene policy negative sample data need to be constructed from one scene policy positive sample data. In different scene tag data, not only the scene attribute identified by the scene tag data may be different, but also the number of the scene attribute identified by the scene tag data may be different.

In step S203, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data are obtained according to the scene policy positive sample data and the scene policy negative sample data.

The first target crowd tag vector, the first target scene tag vector and the first target object tag vector are tag vectors obtained by mapping the first crowd tag data, the first target scene tag data and the first target object tag data into a multi-dimensional vector in sequence. Wherein, the multi-dimension is generally 64-dimension. In the first embodiment of the application, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data are obtained according to scene policy positive sample data and scene policy negative sample data.

Firstly, a first initial crowd label vector corresponding to first crowd label data is obtained, a first initial scene label vector corresponding to first scene label data is obtained, and a first initial object label vector corresponding to first object label data is obtained. In the first embodiment of the present application, please refer to fig. 3, which is a schematic diagram of a tag vector obtaining method in the first embodiment of the present application.

The implementation manner of obtaining the first initial population label vector is as follows: each first-crowd sub-tag data included in the first-crowd tag data is obtained. A first initial population sub-label vector corresponding to each first population sub-label data included in the first population label data is obtained. And carrying out weighted average processing on all the first initial population sub-label vectors to obtain first initial population label vectors.

The number of at least one of the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector is plural.

The first-person group tag data is taken as: for example, the first-person group tag data includes four first-person group sub-tag data for each of the male, middle-aged, white-collar, and married persons, which are in turn: first sub-tag data: male, second sub-tag data: middle year, third sub-label data: white collar, fourth sub-label data: the marriage is already done. In this case, the process of obtaining the first initial population sub-label vectors corresponding to each of the first population sub-label data included in the first population label data is to map the male, the middle year, the white neck, and the married into four first initial population sub-label vectors, that is, the first initial population sub-label vector corresponding to the male, the first initial population sub-label vector corresponding to the middle year, the first initial population sub-label vector corresponding to the white neck, and the first initial population sub-label vector corresponding to the married, respectively, by using the first vector model 301. Correspondingly, all the first initial population sub-label vectors are subjected to weighted average processing to obtain a first initial population label vector corresponding to a formula: the first initial population tag vector = (first initial population sub-tag vector corresponding to male gender, first preset weight + first initial population sub-tag vector corresponding to middle age, second preset weight + first initial population sub-tag vector corresponding to white collar, third preset weight + first initial population sub-tag vector corresponding to married gender, fourth preset weight)/4. It should be noted that the first vector model 301, the second vector model 302, and the third vector model 303 in fig. 3 are models trained in advance for mapping tag data to tag vectors.

The implementation manner of obtaining the first initial scene tag vector is as follows: each first scene sub-label data included in the first scene label data is obtained. A first initial scene sub-label vector corresponding to each first scene sub-label data included in the first scene label data is obtained. And carrying out weighted average processing on all the first initial scene sub-label vectors to obtain first initial scene label vectors. The specific process is similar to the process of obtaining the first initial population tag vector, and is not described in detail herein. The implementation manner of obtaining the first initial object tag vector is as follows: each first object sub-tag data included in the first object tag data is obtained. Obtaining a first initial object sub-label vector corresponding to each first object sub-label data included in the first object label data. And carrying out weighted average processing on all the first initial object sub-label vectors to obtain first initial object label vectors. The specific process is similar to the process of obtaining the first initial population tag vector, and is not described in detail herein.

And secondly, obtaining a second initial crowd label vector corresponding to the second crowd label data, obtaining a second initial scene label vector corresponding to the second scene label data, and obtaining a second initial object label vector corresponding to the second object label data. And finally, obtaining a second initial crowd label vector corresponding to the second crowd label data, obtaining a second initial scene label vector corresponding to the second scene label data, and obtaining a second initial object label vector corresponding to the second object label data.

The implementation manner of obtaining the second initial population label vector is as follows: obtaining each second population sub-label data comprised by the second population label data. A second initial population sub-label vector corresponding to each second population sub-label data included in the second population label data is obtained. And carrying out weighted average processing on all the second initial population sub-label vectors to obtain second initial population label vectors. The number of at least one of the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector is plural.

And taking the second crowd label data as: for example, the second population tag data includes four second population sub-tag data for each young year, student, and unmarried year, and in this order: first sub-tag data: young, second sub-tag data: student, third sub-label data: and (5) leaving unmarried. In this case, the process of obtaining the second initial population sub-label vectors corresponding to each second population sub-label data included in the second population label data is to map the teenager, the student, and the maid into three second initial population sub-label vectors, that is, the second initial population sub-label vector corresponding to the teenager, the second initial population sub-label vector corresponding to the student, and the second initial population sub-label vector corresponding to the maid, respectively, by using the first vector model 301. Correspondingly, all the second initial population sub-label vectors are subjected to weighted average processing to obtain a second initial population label vector corresponding to a formula: a second initial population label vector = (a second initial population sub-label vector corresponding to youth × first preset weight + a second initial population sub-label vector corresponding to student × second preset weight + a second initial population sub-label vector corresponding to maiden/third preset weight)/3.

The implementation manner of obtaining the second initial scene tag vector is as follows: each second scene sub-label data included in the second scene label data is obtained. And obtaining a second initial scene sub-label vector corresponding to each second scene sub-label data included in the second scene label data. And carrying out weighted average processing on all the second initial scene sub-label vectors to obtain second initial scene label vectors. The specific process is similar to the process of obtaining the second initial population tag vector, and is not described in detail herein. The implementation manner of obtaining the second initial object tag vector is as follows: each second object sub-label data included in the second object label data is obtained. And obtaining a second initial object sub-label vector corresponding to each second object sub-label data included in the second object label data. And carrying out weighted average processing on all the second initial object sub-label vectors to obtain second initial object label vectors. The specific process is similar to the process of obtaining the second initial population tag vector, and is not described in detail herein.

And thirdly, obtaining a first target crowd label vector, a first target scene label vector and a first target object label vector according to the first initial crowd label vector, the first initial scene label vector, the first initial object label vector, the second initial crowd label vector, the second initial scene label vector and the second initial object label vector.

In the first embodiment of the present application, for each first target crowd tag vector, first target scene tag vector, and first target object tag vector, a specific implementation process for obtaining the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector is as follows: firstly, a first initial crowd label vector, a first initial scene label vector, a first initial object label vector, a second initial crowd label vector, a second initial scene label vector and a second initial object label vector are used as input parameters of a loss function, and the first initial crowd label vector, the first initial scene label vector and the first initial object label vector when a loss function value is a minimum value are obtained. Then, the first initial crowd tag vector, the first initial scene tag vector and the first initial object tag vector when the loss function value is the minimum value are determined as a first target crowd tag vector, a first target scene tag vector and a first target object tag vector.

In the first embodiment of the present application, the formula corresponding to the loss function is:

. Wherein h, r and t are a first initial crowd label vector, a first initial scene label vector and a first initial object label vector in sequence,

a first modulo result of a first distance value between a first initial population tag vector, a first initial scene tag vector, and a first initial object tag vector,

and obtaining a second modulus result of a second distance value between a second initial crowd label vector and a second initial scene label vector and a second initial object label vector, wherein margin is a hinge function value. The so-called hinge function being presetFor use in making

Functions that do not exceed a specified value. Taking the first initial crowd tag vector, the first initial scene tag vector and the first initial object tag vector as an example, relative to five second crowd tag data, second scene tag data and second object tag data,

i =1, 2, 3, 4, 5, i.e. LOSS is 0 and

、

、

、

and

the maximum value therebetween.

The corresponding formula of the loss function is

Therefore, taking the first initial population label vector, the first initial scene label vector, the first initial object label vector, the second initial population label vector, the second initial scene label vector, and the second initial object label vector as input parameters of the loss function, the process of obtaining the first initial population label vector, the first initial scene label vector, and the first initial object label vector when the loss function value is the minimum value is as follows:

firstly, a first initial crowd label vector, a first initial scene label vector and a first initial object label vector are used as input parameters of a loss function, and a first initial object label vector is calculatedA first modulo result of a first distance value between the crowd tag vector, the first initial scene tag vector, and the first initial object tag vector. I.e. calculating

. So-called

The calculation formula of (2) is as follows:

that is, the process of calculating the first modulo result of the first distance value between the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector is as follows: first, a first sum value between a first initial crowd tag vector and a first initial scene tag vector is calculated. Then, a first difference between the first sum and the first initial object tag vector is calculated. Finally, a first modulo result for the first difference is obtained.

Secondly, calculating a second initial crowd label vector, a second initial scene label vector and a second initial object label vector as input parameters of a loss function

. So-called

The calculation formula of (2) is as follows:

that is, the process of calculating the second modulo result of the second distance value between the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector is as follows: first, a second sum between a second initial crowd tag vector and a second initial scene tag vector is calculated. Then, a second difference between the second sum and the second initial object tag vector is calculated. Finally, a second modulo result for the second difference is obtained.

And thirdly, calculating a modulus result difference value between the first modulus result and the second modulus result. That is to say that the first and second electrodes,

。

and fourthly, calculating a third sum between the modulus-taking result difference value and the hinge function value. That is to say that the first and second electrodes,

、

、

、

and

。

fifth, the maximum value of the third sum value compared to the preset loss function value is determined as the loss function value. The predetermined loss function value is 0, but may be other predetermined positive values. That is, each time the third sum is obtained, the third sum is used

、

、

、

And

comparing with 0, obtaining the maximum value of the third sum value compared with the preset loss function value, and determining the maximum value as the loss functionThe value is obtained.

And sixthly, obtaining a first initial crowd label vector, a first initial scene label vector and a first initial object label vector corresponding to the minimum value of the loss function.

In order to obtain a first initial population label vector, a first initial scene label vector and a first initial object label vector corresponding to a loss function minimum value, the first initial population label vector, the first initial scene label vector, the second initial object label vector, the second initial population label vector, the second initial scene label vector and the second initial object label vector need to be adjusted through gradient descent continuously until obtaining a loss function minimum value, and at this time, the first initial population label vector, the first initial scene label vector and the first initial object label vector when a loss function value is a minimum value are determined as a first target population label vector, A first target scene tag vector and a first target object tag vector.

In the first embodiment of the present application, when obtaining the first initial population tag vector, the first initial scene tag vector, and the first initial object tag vector corresponding to the minimum value of the loss function by using gradient descent, the formula for solving the gradient is as follows:

wherein, in the step (A),

、

is a preset fixed parameter.

In step S204, preference degree data of the crowd corresponding to the first crowd tag data for the first object tag data in the first scene corresponding to the first scene tag data is obtained according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector.

In the embodiment of the application, after the first target crowd tag vector, the first target scene tag vector and the first target object tag vector are calculated, a third modulus result of a third distance value between the first target crowd tag vector, the first target scene tag vector and the first target object tag vector is calculated, and preference degree data of a crowd corresponding to the first crowd tag data for the first object tag data in a first scene corresponding to the first scene tag data is obtained according to the third modulus result. A third modulo result of calculating a third distance value between the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector takes the following formula: third modulo result =

And when the first target crowd tag vector and the first target scene tag vector are the same and the first target object tag vector is different, the larger the obtained third modulus result is, the larger preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data is.

In the first embodiment of the application, preference degrees of the crowd corresponding to the first crowd tag data to a plurality of different first object tag data in the first scene corresponding to the first scene tag data may be obtained, and the crowd corresponding to the first crowd tag data is sorted to the plurality of different first object tag data in the first scene corresponding to the first scene tag data according to the preference degree data, so as to obtain a preference degree list of the crowd corresponding to the first crowd tag data to the plurality of different first object tag data in the first scene corresponding to the first scene tag data. As shown in fig. 4, it is a table diagram of preference degrees provided in the first embodiment of the present application. Fig. 4 shows that the first-person group tag data is a medium-aged, married, or white-collar group, and the first scene tag data is: a preference degree list composed of preference degree data for a plurality of different first object tag data under city, weekday, time-lunch.

In the first embodiment of the present application, the preference degree data for the specified object tag data is data indicating the preference degree for the specified object tag data, and specifically, may be a selection probability for the specified object tag data.

The preference degree data obtaining method provided by the first embodiment of the application includes the steps that firstly, scene strategy positive sample data is obtained, wherein the scene strategy positive sample data comprises first crowd label data, first scene label data and first object label data with historical association; secondly, constructing scene strategy negative sample data relative to the scene strategy positive sample data, wherein the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation; and thirdly, according to the scene strategy positive sample data and the scene strategy negative sample data, obtaining a first target crowd label vector corresponding to the first crowd label data, a first target scene label vector corresponding to the first scene label data and a first target object label vector corresponding to the first object label data. And finally, according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector, obtaining preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data. According to the preference degree data obtaining method provided by the first embodiment of the application, preference degree data of a crowd corresponding to first crowd tag data to first object tag data in a first scene corresponding to first scene tag data can be obtained, and therefore the problem of how to obtain preference degree data of a user to an object is solved. In addition, preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data is obtained, and the object conforming to the preference of the user can be obtained according to the preference degree data.

Second embodiment

The second embodiment of the present application also provides a preference degree data obtaining apparatus, corresponding to the embodiments corresponding to the application scenarios of the preference degree data obtaining method provided by the present application and the preference degree data obtaining method provided by the first embodiment. Since the embodiment of the apparatus is basically similar to the embodiment corresponding to the application scenario and the first embodiment, the description is relatively simple, and for relevant points, reference may be made to the embodiment corresponding to the application scenario and part of the description of the first embodiment. The device embodiments described below are merely illustrative.

Please refer to fig. 5, which is a schematic diagram of a preference level data obtaining apparatus according to a second embodiment of the present application.

The preference degree data obtaining apparatus provided in the second embodiment of the present application includes: a positive sample data obtaining unit 501, configured to obtain scene policy positive sample data, where the scene policy positive sample data includes first crowd tag data, first scene tag data, and first object tag data having a historical association relationship; a negative sample data constructing unit 502, configured to construct scene policy negative sample data opposite to the scene policy positive sample data, where the scene policy negative sample data includes second crowd tag data, second scene tag data, and second object tag data that do not have a historical association relationship; a tag vector obtaining unit 503, configured to obtain, according to the scene policy positive sample data and the scene policy negative sample data, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data; a preference degree obtaining unit 504, configured to obtain, according to the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector, preference degree data of a crowd corresponding to the first crowd tag data for the first object tag data in a first scene corresponding to the first scene tag data.

Optionally, the positive sample data obtaining unit 501 is specifically configured to obtain a scene policy sample whose number ratio exceeds a preset ratio threshold in the full-scale scene policy sample data, and determine the scene policy sample whose number ratio exceeds the preset ratio threshold as the scene policy positive sample data.

Optionally, the positive sample data obtaining unit 501 is further configured to obtain user historical behavior data;

extracting user data, scene data and object data from the user historical behavior data; obtaining crowd label data corresponding to the user data; obtaining scene tag data corresponding to the scene data;

obtaining object tag data corresponding to the object data; generating scene strategy sample data according to the crowd label data, the scene label data and the object label data; and determining all generated scene strategy sample data as the full-scale scene strategy sample data.

Optionally, the positive sample data obtaining unit 501 is further configured to obtain a plurality of crowd tag data; obtaining a plurality of scene tag data; obtaining a plurality of object tag data; selecting one crowd tag data from the plurality of crowd tag data, selecting one scene tag data from the plurality of scene tag data, and selecting one object tag data from the plurality of object tag data; generating scene strategy sample data according to randomly selected crowd label data, randomly selected scene label data and randomly selected object label data; determining all generated scene strategy sample data as the full-scale scene strategy sample data; and the scene strategy sample with the number ratio exceeding a preset ratio threshold is a scene strategy sample generated according to the historical behavior data of the user.

Optionally, the first scene tag data is the same as the second scene tag data, the first object tag data is the same as the second object tag data, and the negative sample data constructing unit 502 is specifically configured to replace the first crowd tag data included in the scene policy positive sample data with the second crowd tag data that is different from the first crowd tag data, so as to obtain the scene policy negative sample data; or, the first crowd tag data is the same as the second crowd tag data, the first scene tag data is the same as the second scene tag data, and the negative sample data constructing unit 502 is specifically configured to replace the first object tag data included in the scene policy positive sample data with the second object tag data that is different from the first object tag data, so as to obtain the scene policy negative sample data; and the negative sample data of any scene strategy is different from the positive sample data of any scene strategy.

Optionally, the tag vector obtaining unit 503 is specifically configured to obtain a first initial crowd tag vector corresponding to the first crowd tag data, obtain a first initial scene tag vector corresponding to the first scene tag data, and obtain a first initial object tag vector corresponding to the first object tag data; obtaining a second initial crowd label vector corresponding to the second crowd label data, obtaining a second initial scene label vector corresponding to the second scene label data, and obtaining a second initial object label vector corresponding to the second object label data; and obtaining the first target crowd label vector, the first target scene label vector and the first target object label vector according to the first initial crowd label vector, the first initial scene label vector, the first initial object label vector, the second initial crowd label vector, the second initial scene label vector and the second initial object label vector.

Optionally, the preference degree obtaining unit 504 is specifically configured to calculate a third modulo result of a third distance value between the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector; and obtaining preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data according to the third modulus result.

Third embodiment

The third embodiment of the present application also provides an object recommendation method, corresponding to the embodiments corresponding to the application scenarios of the preference degree data obtaining method provided by the present application and the preference degree data obtaining method provided by the first embodiment. Since the third embodiment is basically similar to the embodiment corresponding to the application scenario and the first embodiment, the description is relatively simple, and for relevant points, reference may be made to the embodiment corresponding to the application scenario and part of the description of the first embodiment. The third embodiment described below is merely illustrative.

Please refer to fig. 6, which is a flowchart illustrating an object recommendation method according to a third embodiment of the present application.

Step S601: and obtaining target user information.

In the third embodiment of the present application, the target user information is generally user account information obtained when a target user registers as a user of an online service platform, and is generally a user ID, which is used for logging in a client of the online service platform. The online service platform includes, but is not limited to, a take-out service platform, a power provider service platform, a short video service platform, and the like. A client is typically an application or software installed on a computing device. In this case, the obtaining of the target user information specifically includes: first, in response to a trigger operation for a target application, target user information for logging in the target application is obtained. And then, obtaining target user information provided by a client corresponding to the user.

Step S602: and obtaining target crowd label data corresponding to the target user information.

In a third embodiment of the present application, when a target user accesses a client, a first request message for requesting to obtain target crowd tag data corresponding to target user information is sent to a server. The server side logs in a target user account of the client side through a target user, and further obtains target crowd label data provided by the server side according to the user account of the target user. The specific implementation manners of obtaining the target crowd label data corresponding to the target user according to the user account of the target user include at least the following two ways:

The crowd tag data is preset tag data. The crowd tag data is tag data for identifying attributes of crowd tag data, the attributes of the crowd identified by the crowd tag data include, but are not limited to, gender, age, work, marital status, academic history, etc. the attributes of the crowd identified by the crowd tag data include at least one of gender, age, work, marital status, academic history, etc. of the crowd, and the attributes of the crowd identified by different crowd tag data are different. The gender, age, job, marital status, academic history, etc. of the population within the population attributes identified by the population label data may be further divided according to daily habits or experiences, such as: the age of the population in the population attributes identified by the population label data may be further divided into children, teenagers, middle-aged people, elderly people, etc., and may be further divided into different age groups: …, 0-8, 9-18, 19-40, and 41-65 years old. The work of the crowd in the crowd attributes identified by the crowd label data can be further divided into white collar, blue collar, gold collar and the like, and can also be further divided into different work types, such as: doctors, teachers, attorneys, attendants, etc. …. The crowd tag data may be: middle aged, married, and aged; the method can also comprise the following steps: young, not married, students, men; the following steps can be also included: 20-25 years old, not married, teacher, woman, this family.

Step S603: target scene tag data is obtained.

In the third embodiment of the present application, the scene tag data is tag data for identifying attributes of the scene tag data, the attributes of the scene identified by the scene tag data include, but are not limited to, a time, a space, an environment, and the like to which the scene belongs, the attributes of the scene identified by the scene tag data include at least one of a time, a space, an environment, and the like to which the scene belongs, and the attributes of the scenes identified by different scene tag data are different. Wherein, the space is generally a spatial position, such as: shanghai, Beijing Haihu district, and Beijing Kogyo-rising district, No. 5 building 3 unit. The time, space, environment, etc. to which the scene belongs in the scene attribute identified by the scene tag data may be further divided according to daily habits or experiences, such as: the time of the scene in the scene attribute may be further divided into morning, noon, afternoon, evening, and the like, or may be further divided into time periods: 0-6 hours, 6-12 hours, 12-18 …. The scene tag data may be: noon, last sea, lunch time; the method can also comprise the following steps: noon, beijing, evening, sunny day; the following steps can be also included: 12-18 hours, Chengdu, at home, and in rain.

When the target user accesses the client, the server also obtains the access time information and the access position information of the target user accessing the client through the access log stored in the client, and determines the corresponding environment information under the current access time information and the access position information according to the access time information and the access position information. The visiting location information may be specific longitude and latitude information, or may be specific geographic location information, such as: and a certain district in a certain city has a No. 5 building unit. It is possible to determine whether the location where the user is located under the current visit time information is home or company from the visit location information. According to the access time information and the access position information, the corresponding environment information under the current access time information and the access position information can be further determined, such as: after the current access address is determined to be a certain area in a certain city according to the direction address information and the current access time is determined to be 12 hours on a certain month and a certain day of a certain year according to the access time information, the weather of the certain area in the certain city on the certain day 12 hours on the certain month and the certain day of the certain year can be further determined, namely, the current access time information and the corresponding environment information under the access position information are obtained.

In a third embodiment of the present application, the target scene tag data includes target time sub-tag data, target space sub-tag data, and target environment sub-tag data, and the second scene tag data includes second time sub-tag data, second space sub-tag data, and second environment sub-tag data. Correspondingly, the specific implementation manner of obtaining the target scene tag data is as follows: at least one of target time data, target space data, and target environment data is first obtained by a running device running a target application. Then, at least one of the following sub-label data is obtained: target time sub-label data corresponding to the target time data, target space sub-label data corresponding to the target space data, and target environment sub-label data corresponding to the target environment data. And finally, generating target scene label data according to the target time sub-label data, the target space sub-label data and the target environment sub-label data.

Step S604: and recommending the target object data corresponding to the target object tag data to the user corresponding to the target user information according to the preference degree data of the crowd corresponding to the first target crowd tag data obtained in advance for the first candidate object tag data in the first target scene corresponding to the first target scene tag data.

After the client side obtains the target crowd label data and the target scene label data, a second request message for requesting to obtain preference degree data of the crowd corresponding to the target crowd label data to the candidate object label data in the target scene corresponding to the target scene label data is sent to the server side. The server side obtains preference degree data of the crowd corresponding to the target crowd tag data provided by the server side for the candidate object tag data in the target scene corresponding to the target scene tag data, aiming at the second request message.

The specific implementation manner of recommending the target object data corresponding to the target object tag data to the user corresponding to the target user information is as follows: and sending the target object data to the client corresponding to the user. The preference degree data of the crowd corresponding to the target object data, namely the target first crowd tag data, in the target first scene corresponding to the first target scene tag data is displayed on the page of the target application, and the preference degree data of the crowd corresponding to the target first crowd tag data in the target first scene corresponding to the first target scene tag data is obtained according to the first target crowd tag vector corresponding to the target first crowd tag data, the first target scene tag vector corresponding to the target first scene tag data and the first target candidate object tag vector corresponding to the first candidate object tag data.

The first target crowd tag vector, the first target scene tag vector and the first target candidate object tag vector are obtained according to scene strategy positive sample data and scene strategy negative sample data. The scene policy positive sample data includes first target crowd tag data, target first scene tag data and first candidate object tag data having a historical association relationship, the scene policy negative sample data includes second crowd tag data, second scene tag data and second object tag data not having a historical association relationship, and the scene policy negative sample data is negative sample data opposite to the scene policy positive sample data.

In the third embodiment of the present application, preference degree data of a crowd corresponding to target crowd tag data to candidate object tag data in a target scene corresponding to target scene tag data, which is obtained in advance, is preference degree data of a crowd corresponding to target crowd tag data to a plurality of different candidate object tag data in a target scene corresponding to target scene tag data. In this case, a specific implementation manner of recommending the target object data corresponding to the target object tag data to the user corresponding to the target user information according to the preference degree data of the crowd corresponding to the target crowd tag data obtained in advance for the candidate object tag data in the target scene corresponding to the target scene tag data may be: obtaining candidate object tag data corresponding to the highest value corresponding to the preference degree data; determining candidate object tag data corresponding to the highest value as target object tag data; obtaining target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

In addition, a specific implementation manner of recommending the target object data corresponding to the target object tag data to the user corresponding to the target user information according to the preference degree data of the crowd corresponding to the target crowd tag data obtained in advance for the candidate object tag data in the target scene corresponding to the target scene tag data may be as follows: obtaining candidate object tag data corresponding to preference degree data exceeding a preset value; determining candidate object tag data corresponding to preference degree data exceeding a preset value as target object tag data; obtaining target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

The third embodiment of the present application provides an object recommendation method, which includes first obtaining target user information; secondly, target crowd label data corresponding to the target user information is obtained; obtaining target scene label data again; finally, recommending target object data corresponding to the target object tag data to a user corresponding to the target user information according to preference degree data of a crowd corresponding to the first target crowd tag data obtained in advance for the first candidate object tag data in a first target scene corresponding to the first target scene tag data; the preference degree data of the crowd corresponding to the target first crowd label data to the first candidate object label data in the target first scene corresponding to the first target scene label data is obtained according to a first target crowd label vector corresponding to the target first crowd label data, a first target scene label vector corresponding to the target first scene label data and a first target candidate object label vector corresponding to the first candidate object label data, the first target crowd label vector, the first target scene label vector and the first target candidate object label vector are obtained according to scene policy positive sample data and scene policy negative sample data, the scene policy positive sample data comprises the first target crowd label data, the target first scene label data and the first candidate object label data which have historical association relation, the scene policy negative sample data comprises the second crowd label data which do not have historical association relation, the first candidate object label data and the first target object label data, Second scene tag data and second object tag data, the scene policy negative sample data being negative sample data as opposed to the scene policy positive sample data. The object recommendation method provided by the third embodiment of the application can obtain the preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data, thereby solving the problem of how to obtain the preference degree data of the user to the object. And obtaining the object which accords with the preference of the user according to the preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data in the first scene corresponding to the first scene tag data.

In addition, the target object data recommended to the user corresponding to the target user information is the target object data corresponding to the determined target object tag data according to the preference degree data of the crowd corresponding to the first target crowd tag data obtained in advance for the first candidate object tag data in the first target scene corresponding to the first target scene tag data. Therefore, the target object is more in line with the target user preference, thereby improving the accuracy of the recommended target object.

Fourth embodiment

Corresponding to the embodiments corresponding to the application scenarios of the object recommendation method provided by the present application and the object recommendation method provided by the third embodiment, the fourth embodiment of the present application further provides an object recommendation method. Since the method embodiment is basically similar to the embodiment corresponding to the application scenario and the third embodiment, the description is relatively simple, and for relevant points, reference may be made to the embodiment corresponding to the application scenario and part of the description of the third embodiment. The method embodiments described below are merely illustrative.

Please refer to fig. 7, which is a diagram illustrating an object recommendation method according to a fourth embodiment of the present application.

An object recommendation apparatus provided in a fourth embodiment of the present application includes: a target user information obtaining unit 701 configured to obtain target user information; a target crowd tag data obtaining unit 702, configured to obtain target crowd tag data corresponding to the target user information; a target scene tag data obtaining unit 703 for obtaining target scene tag data; a target object data recommending unit 704, configured to recommend, to a user corresponding to the target user information, target object data corresponding to target object tag data according to pre-obtained preference degree data of a crowd corresponding to first target crowd tag data in a first target scene corresponding to the first target scene tag data; wherein preference degree data of a crowd corresponding to the target first-person group tag data for first candidate object tag data in a target first scene corresponding to first target scene tag data is obtained according to a first target first-person group tag vector corresponding to the target first-person group tag data, a first target scene tag vector corresponding to the target first-person group tag data, and a first target candidate object tag vector corresponding to the first candidate object tag data, the first target crowd tag vector, the first target scene tag vector, and the first target candidate object tag vector being obtained according to scene policy positive sample data and scene policy negative sample data, the scene policy positive sample data including the first target crowd tag data, the target first scene tag data, and the first candidate object tag data having a historical association relationship, the scene policy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation, and the scene policy negative sample data is negative sample data opposite to the scene policy positive sample data.

Optionally, the target object data recommending unit 704 is specifically configured to obtain candidate object tag data corresponding to a highest value corresponding to the preference degree data; determining candidate object tag data corresponding to the highest value as the target object tag data; obtaining the target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

Optionally, the target object data recommending unit 704 is specifically configured to obtain candidate object tag data corresponding to preference degree data exceeding a preset value; determining candidate object tag data corresponding to preference degree data exceeding a preset value as the target object tag data; obtaining the target object data corresponding to the target object tag data; and recommending the target object data to the user corresponding to the target user information.

Optionally, the target user information obtaining unit 701 is specifically configured to obtain the target user information, and includes: obtaining the target user information provided by a client corresponding to the user; the target object data recommending unit 704 is specifically configured to send the target object data to the client corresponding to the user.

Optionally, the target user information obtaining unit 701 is specifically configured to obtain, in response to a trigger operation for a target application, target user information used for logging in the target application; the target object data recommending unit 704 is specifically configured to display the target object data on a page of the target application.

Optionally, the target crowd tag data obtaining unit 702 is specifically configured to send, to a server, a first request message for requesting to obtain target crowd tag data corresponding to the target user information; and obtaining the target crowd label data provided by the server.

Optionally, the target scene tag data obtaining unit 703 is specifically configured to obtain at least one of target time data, target spatial data, and target environment data through a running device running the target application; obtaining at least one of the following seed label data: target time sub-label data corresponding to the target time data, target space sub-label data corresponding to the target space data, and target environment sub-label data corresponding to the target environment data; and generating the target scene label data according to the target time sub-label data, the target space sub-label data and the target environment sub-label data.

Optionally, the object recommending apparatus provided in the fourth embodiment of the present application further includes: a second request information sending unit, configured to send, to a server, a second request message for requesting to obtain preference degree data of a crowd corresponding to the target crowd tag data to candidate object tag data in a target scene corresponding to the target scene tag data; and the preference degree data obtaining unit is used for obtaining preference degree data of the crowd corresponding to the target crowd label data provided by the server side for the candidate object label data in the target scene corresponding to the target scene label data.

Fifth embodiment

Corresponding to the above method embodiments provided by the present application, a fifth embodiment of the present application further provides an electronic device. Since the fifth embodiment is substantially similar to the above method embodiment provided in this application, the description is relatively simple, and the relevant points can be referred to the partial description of the above method embodiment provided in this application. The fifth embodiment described below is merely illustrative.

Please refer to fig. 8, which is a schematic diagram of an electronic device provided in an embodiment of the present application.

The electronic device includes: a processor 801; and a memory 802 for storing a program of an obtaining method of distribution pressure data, the apparatus being powered on and executing the program of the obtaining method of distribution pressure data by the processor to perform the method provided in the above-described embodiment of the present application.

It should be noted that, for the detailed description of the electronic device provided in the fifth embodiment of the present application, reference may be made to the related description of the foregoing method embodiment provided in the present application, and details are not repeated here.

Sixth embodiment

Corresponding to the above method embodiments provided by the present application, a sixth embodiment of the present application further provides a storage medium. Since the sixth embodiment is substantially similar to the above method embodiment provided in this application, the description is relatively simple, and for the relevant points, reference may be made to part of the description of the above method embodiment provided in this application. The sixth embodiment described below is merely illustrative.

The storage medium stores a computer program that is executed by a processor to perform the methods provided in the above-described embodiments of the present application.

It should be noted that, for the detailed description of the storage medium provided in the sixth embodiment of the present application, reference may be made to the related description of the foregoing method embodiment provided in the present application, and details are not repeated here.

Although the present invention has been described with reference to the preferred embodiments, it should be understood that the scope of the present invention is not limited to the embodiments described above, and that various changes and modifications may be made by one skilled in the art without departing from the spirit and scope of the present invention.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or Flash memory (Flash RAM). Memory is an example of a computer-readable medium.

1. Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable Media does not include non-Transitory computer readable Media (transient Media), such as modulated data signals and carrier waves.

2. As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

Claims

1. A preference degree data obtaining method, comprising:

obtaining scene strategy positive sample data, wherein the scene strategy positive sample data comprises first crowd label data, first scene label data and first object label data with historical incidence relation;

constructing scene strategy negative sample data opposite to the scene strategy positive sample data, wherein the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation;

according to the scene strategy positive sample data and the scene strategy negative sample data, obtaining a first target crowd label vector corresponding to the first crowd label data, a first target scene label vector corresponding to the first scene label data and a first target object label vector corresponding to the first object label data;

according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector, preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data is obtained.

2. The method of claim 1, wherein obtaining scene policy positive sample data comprises: obtaining a scene strategy sample with the quantity ratio exceeding a preset ratio threshold value in the full-scale scene strategy sample data, and determining the scene strategy sample with the quantity ratio exceeding the preset ratio threshold value as the scene strategy positive sample data.

3. The method of claim 2, further comprising:

obtaining historical behavior data of a user;

extracting user data, scene data and object data from the user historical behavior data;

obtaining crowd label data corresponding to the user data;

obtaining scene tag data corresponding to the scene data;

obtaining object tag data corresponding to the object data;

generating scene strategy sample data according to the crowd label data, the scene label data and the object label data;

and determining all generated scene strategy sample data as the full-scale scene strategy sample data.

4. The method of claim 2, further comprising:

obtaining a plurality of crowd tag data;

obtaining a plurality of scene tag data;

obtaining a plurality of object tag data;

selecting one crowd tag data from the plurality of crowd tag data, selecting one scene tag data from the plurality of scene tag data, and selecting one object tag data from the plurality of object tag data;

generating scene strategy sample data according to randomly selected crowd label data, randomly selected scene label data and randomly selected object label data;

determining all generated scene strategy sample data as the full-scale scene strategy sample data;

and the scene strategy sample with the number ratio exceeding a preset ratio threshold is a scene strategy sample generated according to the historical behavior data of the user.

5. The method of claim 1, wherein the first scene tag data is the same as the second scene tag data, wherein the first object tag data is the same as the second object tag data, and wherein constructing scene policy negative sample data as opposed to the scene policy positive sample data comprises: replacing the first crowd tag data included in the scene policy positive sample data with the second crowd tag data different from the first crowd tag data to obtain the scene policy negative sample data; or, the first crowd tag data is the same as the second crowd tag data, the first scene tag data is the same as the second scene tag data, and the constructing of scene policy negative sample data relative to the scene policy positive sample data includes: replacing the first object tag data included in the scene policy positive sample data with the second object tag data different from the first object tag data to obtain the scene policy negative sample data;

and the negative sample data of any scene strategy is different from the positive sample data of any scene strategy.

6. The method according to claim 1, wherein the obtaining a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data according to the scene policy positive sample data and the scene policy negative sample data comprises:

obtaining a first initial crowd label vector corresponding to the first crowd label data, obtaining a first initial scene label vector corresponding to the first scene label data, and obtaining a first initial object label vector corresponding to the first object label data;

obtaining a second initial crowd label vector corresponding to the second crowd label data, obtaining a second initial scene label vector corresponding to the second scene label data, and obtaining a second initial object label vector corresponding to the second object label data;

and obtaining the first target crowd label vector, the first target scene label vector and the first target object label vector according to the first initial crowd label vector, the first initial scene label vector, the first initial object label vector, the second initial crowd label vector, the second initial scene label vector and the second initial object label vector.

7. The method of claim 6, wherein obtaining a first initial population tag vector corresponding to the first population tag data comprises: obtaining a first initial population sub-label vector corresponding to each first population sub-label data included in the first population label data, and performing weighted average processing on all the first initial population sub-label vectors to obtain the first initial population label vectors;

or, the obtaining a first initial scene tag vector corresponding to the first scene tag data includes: obtaining a first initial scene sub-label vector corresponding to each first scene sub-label data included in the first scene label data, and performing weighted average processing on all the first initial scene sub-label vectors to obtain the first initial scene label vectors;

or, the obtaining a first initial object tag vector corresponding to the first object tag data includes: obtaining a first initial object sub-label vector corresponding to each first object sub-label data included in the first object label data, and performing weighted average processing on all the first initial object sub-label vectors to obtain the first initial object label vectors;

or, the obtaining a second initial crowd tag vector corresponding to the second crowd tag data includes: obtaining a second initial population sub-label vector corresponding to each second population sub-label data included in the second population label data, and performing weighted average processing on all second initial population sub-label vectors to obtain a second initial population label vector;

or, the obtaining a second initial scene tag vector corresponding to the second scene tag data includes: obtaining a second initial scene sub-label vector corresponding to each second scene sub-label data included in the second scene label data, and performing weighted average processing on all the second initial scene sub-label vectors to obtain a second initial scene label vector;

or, the obtaining a second initial object tag vector corresponding to the second object tag data includes: and obtaining a second initial object sub-label vector corresponding to each second object sub-label data included in the second object label data, and performing weighted average processing on all the second initial object sub-label vectors to obtain the second initial object label vectors.

8. The method of claim 6, wherein the number of at least one of the first initial crowd tag vector, the first initial scene tag vector, the first initial object tag vector is plural;

the obtaining the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector according to the first initial crowd tag vector, the first initial scene tag vector, the first initial object tag vector, the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector includes:

taking the first initial population label vector, the first initial scene label vector, the first initial object label vector, the second initial population label vector, the second initial scene label vector and the second initial object label vector as input parameters of a loss function, and obtaining the first initial population label vector, the first initial scene label vector and the first initial object label vector when a loss function value is a minimum value;

and determining a first initial crowd tag vector, a first initial scene tag vector and a first initial object tag vector when the loss function value is a minimum value as the first target crowd tag vector, the first target scene tag vector and the first target object tag vector.

9. The method of claim 8, wherein the step of obtaining the first initial crowd tag vector, the first initial scene tag vector, the first initial object tag vector, the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector with the loss function value at the minimum value as input parameters of the loss function comprises:

calculating a first modulo result of a first distance value between the first initial crowd tag vector, the first initial scene tag vector, and the first initial object tag vector;

calculating a second modulo result of a second distance value between the second initial crowd tag vector, the second initial scene tag vector, and the second initial object tag vector;

calculating a modulus result difference between the first modulus result and the second modulus result;

calculating a third sum between the modulus-taking result difference and the hinge function value;

determining a maximum value of the third sum value compared to a preset loss function value as a loss function value;

determining the minimum value of all the obtained loss function values as the minimum value of the loss function;

and obtaining a first initial crowd label vector, a first initial scene label vector and a first initial object label vector corresponding to the minimum value of the loss function.

10. The method of claim 9, wherein said computing a first modulo result of a first distance value between said first initial population tag vector, said first initial scene tag vector, and said first initial object tag vector comprises:

calculating a first sum between the first initial crowd tag vector and the first initial scene tag vector;

calculating a first difference between the first sum and the first initial object tag vector;

a first modulo result for the first difference is obtained.

11. The method of claim 9, wherein said computing a second modulo result of a second distance value between said second initial crowd tag vector, said second initial scene tag vector, and said second initial object tag vector comprises:

calculating a second sum between the second initial crowd tag vector and the second initial scene tag vector;

calculating a second difference between the second sum and the second initial object tag vector;

obtaining a second modulo result for the second difference.

12. The method according to claim 1, wherein the obtaining preference degree data of the crowd corresponding to the first crowd tag data for the first object tag data under the first scene corresponding to the first scene tag data according to the first target crowd tag vector, the first target scene tag vector and the first target object tag vector comprises:

calculating a third modulo result of a third distance value between the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector;

and obtaining preference degree data of the crowd corresponding to the first crowd tag data to the first object tag data under the first scene corresponding to the first scene tag data according to the third modulus result.

13. The method of claim 1, wherein the first scene tag data comprises first temporal sub-tag data, first spatial sub-tag data, and first environmental sub-tag data;

or, the second scene tag data includes second temporal sub-tag data, second spatial sub-tag data, and second environmental sub-tag data.

14. A preference degree data obtaining apparatus, comprising:

the system comprises a positive sample data obtaining unit, a positive sample data obtaining unit and a negative sample data obtaining unit, wherein the positive sample data obtaining unit is used for obtaining scene strategy positive sample data which comprises first crowd tag data, first scene tag data and first object tag data with historical incidence relation;

the negative sample data construction unit is used for constructing scene strategy negative sample data relative to the scene strategy positive sample data, and the scene strategy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have a historical incidence relation;

a tag vector obtaining unit, configured to obtain, according to the scene policy positive sample data and the scene policy negative sample data, a first target crowd tag vector corresponding to the first crowd tag data, a first target scene tag vector corresponding to the first scene tag data, and a first target object tag vector corresponding to the first object tag data;

a preference degree obtaining unit, configured to obtain, according to the first target crowd tag vector, the first target scene tag vector, and the first target object tag vector, preference degree data of a crowd corresponding to the first crowd tag data for the first object tag data in a first scene corresponding to the first scene tag data.

15. An object recommendation method, comprising:

obtaining target user information;

obtaining target crowd label data corresponding to the target user information;

obtaining target scene label data;

recommending target object data corresponding to target object tag data to a user corresponding to the target user information according to pre-obtained preference degree data of a crowd corresponding to the first target crowd tag data in a first target scene corresponding to the first target scene tag data;

wherein preference degree data of a crowd corresponding to the target first-person group tag data for first candidate object tag data in a target first scene corresponding to first target scene tag data is obtained according to a first target first-person group tag vector corresponding to the target first-person group tag data, a first target scene tag vector corresponding to the target first-person group tag data, and a first target candidate object tag vector corresponding to the first candidate object tag data, the first target crowd tag vector, the first target scene tag vector, and the first target candidate object tag vector being obtained according to scene policy positive sample data and scene policy negative sample data, the scene policy positive sample data including the first target crowd tag data, the target first scene tag data, and the first candidate object tag data having a historical association relationship, the scene policy negative sample data comprises second crowd label data, second scene label data and second object label data which do not have historical incidence relation, and the scene policy negative sample data is negative sample data opposite to the scene policy positive sample data.

16. The method according to claim 15, wherein recommending, to the user corresponding to the target user information, the target object data corresponding to the target object tag data according to the preference degree data of the crowd corresponding to the target crowd tag data obtained in advance for the candidate object tag data in the target scene corresponding to the target scene tag data comprises:

obtaining candidate object tag data corresponding to the highest value corresponding to the preference degree data;

determining candidate object tag data corresponding to the highest value as the target object tag data;

obtaining the target object data corresponding to the target object tag data;

and recommending the target object data to the user corresponding to the target user information.

17. The method according to claim 15, wherein recommending, to the user corresponding to the target user information, the target object data corresponding to the target object tag data according to the preference degree data of the crowd corresponding to the target crowd tag data obtained in advance for the candidate object tag data in the target scene corresponding to the target scene tag data comprises:

obtaining candidate object tag data corresponding to preference degree data exceeding a preset value;

determining candidate object tag data corresponding to preference degree data exceeding a preset value as the target object tag data;

obtaining the target object data corresponding to the target object tag data;

18. The method of claim 15, wherein obtaining target user information comprises: obtaining the target user information provided by a client corresponding to the user;

the recommending target object data corresponding to the target object tag data to the user corresponding to the target user information includes: and sending the target object data to a client corresponding to the user.

19. The method of claim 15, wherein obtaining target user information comprises: responding to a trigger operation aiming at a target application, and obtaining target user information for logging in the target application;

the recommending target object data corresponding to the target object tag data to the user corresponding to the target user information includes: and displaying the target object data on the page of the target application.

20. The method according to claim 15 or 19, wherein the obtaining target crowd label data corresponding to the target user information comprises:

sending a first request message for requesting to obtain target crowd label data corresponding to the target user information to a server;

and obtaining the target crowd label data provided by the server.

21. The method of claim 19, wherein obtaining target scene tag data comprises:

obtaining at least one of target time data, target space data and target environment data through a running device running the target application;

obtaining at least one of the following seed label data: target time sub-label data corresponding to the target time data, target space sub-label data corresponding to the target space data, and target environment sub-label data corresponding to the target environment data;

and generating the target scene label data according to the target time sub-label data, the target space sub-label data and the target environment sub-label data.

22. The method of claim 15 or 19, further comprising:

sending a second request message for requesting to obtain preference degree data of the crowd corresponding to the target crowd tag data to the candidate object tag data in the target scene corresponding to the target scene tag data to a server;

and obtaining preference degree data of the crowd corresponding to the target crowd label data provided by the server side for candidate object label data in the target scene corresponding to the target scene label data.

23. An object recommendation apparatus, characterized in that,

a target user information obtaining unit for obtaining target user information;

the target crowd tag data obtaining unit is used for obtaining target crowd tag data corresponding to the target user information;

a target scene tag data obtaining unit for obtaining target scene tag data;

the target object data recommending unit is used for recommending target object data corresponding to the target object tag data to a user corresponding to the target user information according to pre-obtained preference degree data of a crowd corresponding to the first target crowd tag data to the first candidate object tag data in a first target scene corresponding to the first target scene tag data;

24. An electronic device, comprising:

a processor; and

memory for storing a program of a method, the device performing the method of any one of claims 1-13 and 15-22 when powered on and running the program of the method via the processor.

25. A storage medium, characterized in that the storage medium stores a computer program which is run by a processor for performing the method of any one of claims 1-13 and 15-22.