CN112418920A - Training method of information recommendation model, information recommendation method and device - Google Patents

Training method of information recommendation model, information recommendation method and device Download PDF

Info

Publication number
CN112418920A
CN112418920A CN202011254218.1A CN202011254218A CN112418920A CN 112418920 A CN112418920 A CN 112418920A CN 202011254218 A CN202011254218 A CN 202011254218A CN 112418920 A CN112418920 A CN 112418920A
Authority
CN
China
Prior art keywords
user
information recommendation
information
recommended
recommendation
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011254218.1A
Other languages
Chinese (zh)
Inventor
廖国钢
吴晓旭
石晓文
崔同
王永康
王兴星
张硕
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Sankuai Online Technology Co Ltd
Original Assignee
Beijing Sankuai Online Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Sankuai Online Technology Co Ltd filed Critical Beijing Sankuai Online Technology Co Ltd
Priority to CN202011254218.1A priority Critical patent/CN112418920A/en
Publication of CN112418920A publication Critical patent/CN112418920A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • G06Q30/0269—Targeted advertisements based on user profile or attribute
    • G06Q30/0271—Personalized advertisement
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/02—Marketing; Price estimation or determination; Fundraising
    • G06Q30/0241—Advertisements
    • G06Q30/0251—Targeted advertisements
    • G06Q30/0255—Targeted advertisements based on user history

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Strategic Management (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Accounting & Taxation (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The specification discloses a training method of an information recommendation model, an information recommendation method and a device, which are used for acquiring historical behavior data of each user when the user browses information on an information recommendation list containing a set object and determining behavior characteristic data of each user. And inputting the behavior characteristic data and the recommendation attribute data corresponding to each candidate set object into an information recommendation model to be trained, and inputting each obtained set object to be recommended into a preset information recommendation simulation system for simulation recommendation so as to determine a recommendation effect representation value corresponding to each set object to be recommended. And determining the reward value of the reward function corresponding to the information recommendation model, and training the information recommendation model. The information recommendation model is used for recommending set objects for the user. The method carries out training of the information recommendation model through behavior characteristic data of the recommendation information of the user, and recommends the information to the user through the trained information recommendation model, so that information sequencing of the information recommendation list is optimized.

Description

Training method of information recommendation model, information recommendation method and device
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a training method, an information recommendation method, and an information recommendation apparatus for an information recommendation model.
Background
Currently, a service platform may recommend various types of advertisements to users to improve the life experience of each user.
In practical application, the service platform displays an information recommendation list containing advertisements and each piece of recommendation information to a user, that is, displays the advertisements and each piece of recommendation information to the user in a mixed manner, wherein the positions of the advertisements in the information recommendation list are usually determined by preset weight parameters. The mixed arrangement mode does not consider the different preference degrees of different users on the advertisements, such as the influence of the advertisement quantity on the sequencing of the recommendation information and the influence of the different advertisement positions on the sequencing of the recommendation information, so that the situation that the sequencing of the recommendation information is unreasonable occurs.
Therefore, how to optimize the ranking of the recommendation information more reasonably is an urgent problem to be solved.
Disclosure of Invention
The present specification provides a training method and apparatus for an information recommendation model, a storage medium, and an electronic device, so as to partially solve the above problems in the prior art.
The technical scheme adopted by the specification is as follows:
the present specification provides a training method of an information recommendation model, including:
acquiring historical behavior data of each user when browsing information of an information recommendation list containing a set object;
for each user, determining behavior characteristic data of the user according to the historical behavior data;
inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained to obtain each setting object to be recommended and a display sequence of each setting object to be recommended;
inputting the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation so as to determine recommendation effect representation values corresponding to the setting objects to be recommended in the display sequence;
and determining an incentive value of an incentive function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the incentive value, wherein the information recommendation model is used for recommending a set object to a user.
Optionally, the historical behavior data includes: the number of the setting objects which are actually clicked and browsed by the user in the setting objects displayed by the information recommendation list, the number of the setting objects which are not clicked and browsed by the user in the setting objects displayed by the information recommendation list, and the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
Optionally, determining the behavior feature data of the user according to the historical behavior data specifically includes:
and determining the loss rate of the set objects of the user aiming at the information recommendation list according to the number of the set objects actually clicked and browsed by the user aiming at the set objects displayed by the information recommendation list and the number of the set objects not clicked and browsed by the user aiming at the set objects displayed by the information recommendation list, wherein the loss rate of the set objects is larger if the number of the set objects not clicked and browsed by the user aiming at the set objects displayed by the information recommendation list is higher.
Optionally, determining the behavior feature data of the user according to the historical behavior data specifically includes:
and determining the information browsing depth of the user according to the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
Optionally, the information recommendation model includes: a weighting function;
inputting the behavior feature data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained to obtain each setting object to be recommended, and the method specifically comprises the following steps:
inputting the behavior characteristic data of the user into a preset weight function, and determining a weight parameter corresponding to the user;
and determining each set object to be recommended output by the information recommendation model to be trained according to the weight parameter corresponding to the user and the recommendation attribute data corresponding to each candidate set object.
Optionally, training the information recommendation model according to the reward value specifically includes:
and training the information recommendation model by adjusting the information recommendation model and model parameters contained in the weight function according to the maximum reward value as an optimization target.
Optionally, determining, according to the recommendation effect characterization value, an incentive value of an incentive function corresponding to the information recommendation model, specifically including:
determining a first influence factor and a second influence factor according to the recommendation effect representation value, wherein the first influence factor is used for representing the influence of the user individuals on the profit of the set object, and the second influence factor is used for controlling the influence of the number of the set objects recommended to the user on the recommendation effect of the user individuals and the set object;
and determining the reward value of the reward function corresponding to the information recommendation model according to the recommendation effect representation value, the first influence factor and the second influence factor.
The present specification provides an information recommendation method, including:
acquiring historical behavior data of a user for an information list when the user browses information;
determining behavior characteristic data corresponding to the user according to the historical behavior data;
acquiring historical behavior data when a user browses information on an information recommendation list containing a set object;
determining behavior characteristic data corresponding to the user according to the historical behavior data;
inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommendation model to obtain each setting object to be recommended and a display sequence of each setting object to be recommended, wherein the information recommendation model is obtained by training through the method;
and recommending the setting objects to be recommended to the user in the information to be recommended in a mixed arrangement mode according to the display sequence.
Optionally, the behavior feature data and recommendation attribute data corresponding to each candidate setting object are input into a pre-trained information recommendation model to obtain each setting object to be recommended and an arrangement order of the setting objects to be recommended, and the method specifically includes:
inputting the behavior characteristic data into a preset weight function, and determining the weight function corresponding to the user, wherein the weight parameters determined by different users through the weight function are not identical;
and determining each set object to be recommended and the arrangement sequence of each set object to be recommended, which are output by the information recommendation model, according to the weight parameters corresponding to the user and the recommendation attribute data corresponding to each candidate set object.
This specification provides a device of information recommendation model training, includes:
the acquisition module is used for historical behavior data of each user when the information recommendation list containing the set object is browsed;
the determining module is used for determining the behavior characteristic data of each user according to the historical behavior data;
the recommendation module is used for inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate set object into an information recommendation model to be trained to obtain each set object to be recommended and a display sequence of each set object to be recommended;
the simulation module is used for inputting the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation so as to determine corresponding recommendation effect representation values of the setting objects to be recommended in the display sequence;
and the training module is used for determining the reward value of the reward function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the reward value, wherein the information recommendation model is used for recommending a set object to a user.
The present specification provides an information recommendation apparatus including:
the acquisition module is used for acquiring historical behavior data when a user browses information on an information recommendation list containing a set object;
the determining module is used for determining behavior characteristic data corresponding to the user according to the historical behavior data;
the recommendation module is used for inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate set object into a pre-trained information recommendation model to obtain each set object to be recommended and a display sequence of each set object to be recommended, and the information recommendation model is obtained by training through the method;
and the pushing module is used for recommending the setting objects to be recommended to the user in the information to be recommended in a mixed arrangement mode according to the display sequence.
The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described information recommendation model training method or the above-described information recommendation method.
The present specification provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the training method of the information recommendation model or the information recommendation method when executing the program.
The technical scheme adopted by the specification can achieve the following beneficial effects:
in the training method of the information recommendation model provided in this specification, historical behavior data of each user when browsing information for information recommendation including a set object is acquired, and for each user, behavior feature data of the user is determined according to the historical behavior data. And then, inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained to obtain each setting object to be recommended and a display sequence of each setting object to be recommended, and inputting each setting object to be recommended and each information to be recommended which are arranged according to the display sequence into a preset information recommendation simulation system for simulation recommendation so as to determine a recommendation effect representation value corresponding to each setting object to be recommended. And finally, determining the reward value of a reward function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the reward value, wherein the information recommendation model is used for recommending set objects to the user.
According to the method, the preference degree of the user to each set object in the information recommendation list can be determined based on the historical behavior data of the user, and the information recommendation model is trained. That is to say, the display position where each setting object in the information recommendation list is located is determined according to the preference of the user, and compared with the prior art in which the display position of each setting object is determined only by a fixed weight parameter, the unreasonable information sequencing of the information recommendation list is avoided, so that the information sequencing of the information recommendation list is optimized.
Drawings
The accompanying drawings, which are included to provide a further understanding of the specification and are incorporated in and constitute a part of this specification, illustrate embodiments of the specification and together with the description serve to explain the specification and not to limit the specification in a non-limiting sense. In the drawings:
fig. 1 is a schematic flowchart of a training method of an information recommendation model provided in an embodiment of the present specification;
fig. 2 is a schematic flowchart of an information recommendation method provided in an embodiment of the present specification;
fig. 3 is a schematic flow chart of information recommendation model training and information recommendation provided in an embodiment of the present disclosure;
fig. 4 is a schematic structural diagram of an information recommendation model training apparatus provided in an embodiment of the present specification;
fig. 5 is a schematic structural diagram of an information recommendation device provided in an embodiment of the present specification;
fig. 6 is a schematic structural diagram of an electronic device provided in an embodiment of this specification.
Detailed Description
In order to make the objects, technical solutions and advantages of the present disclosure more clear, the technical solutions of the present disclosure will be clearly and completely described below with reference to the specific embodiments of the present disclosure and the accompanying drawings. It is to be understood that the embodiments described are only a few embodiments of the present disclosure, and not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort belong to the protection scope of the present specification.
The technical solutions provided by the embodiments of the present description are described in detail below with reference to the accompanying drawings.
In the embodiment of the present specification, before information recommendation is performed on historical behavior data of a user, a pre-trained information recommendation model needs to be relied on, and all that is described below is how to train the information recommendation model, as shown in fig. 1.
Fig. 1 is a schematic flow chart of a training method of an information recommendation model provided in an embodiment of the present specification, which specifically includes the following steps:
s100: and acquiring historical behavior data of each user for the information list when browsing the information.
When a user browses an information recommendation list containing a set object, various historical behavior data can be generated, the historical behavior data can be used as behavior characteristic data for analyzing the information recommendation list browsed by the user, and based on the behavior characteristic data, a service platform can acquire the historical behavior data when each user browses the information recommendation list containing the set object. The setting object mentioned in the present specification may include an advertisement, coupon pickup information, and the like. For convenience of explanation, the method provided in the present specification will be described below mainly in the case where the setting target is an advertisement.
The acquired historical behavior data may include: the number of the set objects actually clicked and browsed by the user in each set object displayed by the information recommendation list (such as the number of advertisements actually clicked and browsed by the user in each advertisement displayed by the information recommendation list), the number of the set objects not clicked and browsed by the user in each set object displayed by the information recommendation list (such as the number of advertisements not clicked and browsed by the user in each advertisement displayed by the information recommendation list), and the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
The number of the set objects actually clicked and browsed by the user in each set object displayed by the information recommendation list is obtained, and the number of the set objects not clicked and browsed by the user in each set object displayed by the information recommendation list can be obtained through statistics respectively. Of course, the service platform may determine the other one of the setting objects when obtaining one of the setting objects, that is, if obtaining the number of the setting objects that the user actually clicks and browses among the setting objects displayed in the information recommendation list, the service platform may directly determine the number of the setting objects that the user does not click and browse among the setting objects displayed in the information recommendation list, and similarly, if obtaining the number of the setting objects that the user does not click and browse among the setting objects displayed in the information recommendation list, the service platform may directly determine the number of the setting objects that the user actually clicks and browses among the setting objects displayed in the information recommendation list.
In the embodiment of the present specification, a plurality of pieces of recommendation information and setting objects are displayed in the information recommendation list, and actually, only a part of recommendation information and setting objects in the information recommendation list may be displayed at the same time on an interface of a terminal device used by a user, and the rest of recommendation information and setting objects need to be displayed by the user performing, for example, a sliding operation on the interface of the terminal device. Based on this, in the embodiment of the present specification, a concept may be introduced, in which the length of the list of the presented part, that is, the length of the list presented by the information recommendation list by the terminal device by performing a specified operation (such as a sliding operation) on the interface of the terminal device when the user browses the information recommendation list. For example, assuming that in a two-column displayed information recommendation list (i.e. one line has two information display bits), the user performs a sliding operation on the interface of the terminal device, so that the terminal device displays 40 information display bits in the information recommendation list on the interface one by one, and the length of the displayed part is 40.
Further, the service platform may count the list length of the displayed part corresponding to each information recommendation list browsed historically by a user, and may further determine the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list historically.
In the embodiment of the present specification, the terminal device used by the user to browse the information may be a terminal device such as a mobile phone, a tablet computer, or the like, and of course, the execution subject used to obtain the information recommendation list may also be a client installed in the terminal device, an Application (App), or the like, or the terminal device or a browser in the client.
The information recommendation list includes, in addition to various setting objects, other information recommended to the user by the service platform, for example, a search result returned to the user by the service platform based on a search keyword input by the user, or a link of each peripheral business actively recommended to the user by the service platform based on the geographical location of the user. Therefore, it should be emphasized that the information recommendation list mentioned in the embodiments of the present specification is obtained by arranging each recommendation information and setting object recommended to the user by the service platform based on a mixed arrangement manner.
S102: and for each user, determining the behavior characteristic data of the user according to the historical behavior data.
In this embodiment of the present specification, each user has corresponding historical behavior data, so the service platform may determine behavior feature data of the user according to the historical behavior data, where the behavior feature data of the user is used to indicate some preference features that the user reflects on a set object when browsing an information recommendation list, and the behavior feature data of the user may include a set object attrition rate, an information browsing depth, an information click preference of the user, and the like.
The set object churn rate can be determined by the number of set objects actually clicked and browsed by the user among the set objects displayed on the information recommendation list and the number of set objects not clicked and browsed by the user among the set objects displayed on the information recommendation list. If the number of the set objects which are not clicked and browsed among the set objects displayed in the information recommendation list by the user is higher, the loss rate of the set objects is higher.
It should be noted that there may be various ways of determining the attrition rate of the setting object by the number of the setting objects actually clicked and browsed by the user among the setting objects displayed in the information recommendation list and the number of the setting objects not clicked and browsed by the user among the setting objects displayed in the information recommendation list. For example, suppose that the service platform historically pushes two information recommendation lists to the user together, and it is determined that both of the two information recommendation lists are browsed by the user according to the number of advertisements actually clicked and browsed by the user in each advertisement (i.e., set object) displayed by the information recommendation list (i.e., set object number) and the number of advertisements not clicked and browsed by the user in each advertisement displayed by the information recommendation list (i.e., set object churn rate), the number of information recommendation lists not browsed by the user is 0, the number of information recommendation lists browsed by the user is 2, and the advertisement churn rate (i.e., set object churn rate) is 0/2 ═ 0, that is, the user browses both information recommendation lists pushed by the service platform, so it is considered that all advertisements included in the information recommendation lists are organically clicked and viewed by the user, and the advertisement churn rate is zero.
For another example, suppose that the service platform historically pushes 10 information recommendation lists to the user, which includes 100 advertisements (i.e., set objects) in the 10 information recommendation lists, and among the 100 advertisements, the user clicks and views 90 advertisements and 10 clicks and views, so that it can be determined that the advertisement loss rate is: 10/100 is 10%.
In this embodiment, the service platform may determine the information browsing depth of the user, that is, the number of information that can be viewed by the user when browsing the information recommendation list once, according to the average list length of the displayed part of the information recommendation list when the user browses each information recommendation list.
For the information click preference of the user, the service platform may determine the information click preference of the user according to the position of the information clicked by the user in each information (including the set object and other information recommended by the service platform) displayed by the information recommendation list, the browsing duration of the clicked information by the user, the information category of the clicked information by the user, and the like.
For example, the service platform may determine, by counting the historical browsing conditions of each information recommendation list by the user, information click preferences of the user, such as whether the information clicked by the user in each information displayed by the information recommendation list is in the front of the list, which information category the user clicks is browsed for is longer, which information category the user clicks is clicked for is more, and the like.
S104: and inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate set object into an information recommendation model to be trained to obtain each set object to be recommended and a display sequence of each set object to be recommended.
In this embodiment of the present specification, the service platform may input the determined behavior feature data and the recommendation attribute data corresponding to each candidate setting object to the information recommendation model to be trained, so as to obtain each setting object to be recommended and a display order of each setting object to be recommended. The recommendation attribute data corresponding to each candidate setting object is used to represent information such as the Cost of the setting object itself and the revenue that the service platform can obtain for the setting object, and for example, when the setting object is an advertisement, the recommendation attribute data may specifically include a Cost Per Mill (CPM) (which also represents the Cost of the advertisement), a Gross transaction Volume (GMV), a commission rate (Takerate), and the like.
Based on this, the service platform may input the recommendation attribute data of the setting object as the feature of the setting object into the information recommendation model to be trained, so as to obtain the setting objects to be recommended and the display order of the setting objects to be recommended. Each setting object to be recommended is screened from a plurality of candidate setting objects, and the screening may be specifically performed by determining a ranking score of the candidate setting object, for example, the ranking score of the candidate setting object may be determined by using the following formula:
RankScore＝f(fk1(user_feature)，fk2(user_feature)，CPM，GMV，Takerate)
wherein f isk1、fk2The method is a weight function in an information recommendation model, the RankScore is used for representing the ranking scores of the candidate set objects, and the user _ feature is the behavior feature data of the user.
It can be seen from the formula that the service platform can input the behavior feature data of the user into a preset weight function, and determine and give the weight parameter corresponding to the user. Then, according to the weight parameters corresponding to the user and the recommendation attribute data (CPM, GMV, Takerate) corresponding to the candidate setting objects, determining the ranking scores of the candidate setting objects, further according to the ranking scores of the candidate setting objects, screening out the setting objects to be recommended from the candidate setting objects, and finally outputting the setting objects to be recommended and the display sequence of the setting objects to be recommended.
It can be seen from the above process that in practical application, behavior feature data corresponding to different users are likely not identical, so that the behavior feature data of different users are input into a preset weight function, and the determined weight parameters are also not identical, so that the display orders of the setting object to be recommended and the setting object to be recommended, which are obtained by different users, are also likely not identical. That is, the weighting parameter corresponding to the user is not constant, and different weighting parameters are obtained according to the optimization of the weighting function or the change of the behavior feature data.
It is to be emphasized that the behavior feature data inputted into the weighting functions may be the same, for example, the above formula actually includes two weighting functions, and the behavior feature data inputted into the two weighting functions are the same. While the behavior characteristic data input to the two weighting functions may be the same, the weighting parameters output by the two weighting functions may be different. In other words, the weighting factors of the two weighting functions are different when the algorithm is designed, so that the output weighting parameters are likely to be different when the same behavior characteristic data is input.
The display sequence mentioned above can be represented by the display position of each to-be-recommended setting object, and this way can not only reflect the arrangement sequence of each to-be-recommended setting object in the information recommendation list displayed to the user in the future, but also reflect the actual position in the information recommendation list, and if it is assumed that the display position corresponding to the to-be-recommended setting object a is 2 and the display position corresponding to the to-be-recommended setting object B is 8, it can be seen that the to-be-recommended setting object a is displayed before the to-be-recommended setting object B, and the information recommendation list also displays 5 pieces of other information between the to-be-recommended setting object a and the to-be-recommended setting object B.
S106: and inputting the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation so as to determine the corresponding recommendation effect representation values of the setting objects to be recommended in the display sequence.
In the embodiment of the present specification, the service platform may input each set object to be recommended and each piece of information to be recommended into a preset information recommendation simulation system for simulation recommendation according to the display sequence. The simulation recommendation comprises a release simulation and a profit simulation, wherein the release simulation is used for performing simulation release on each set object to be recommended and each piece of information to be recommended so as to determine browsing conditions of the user on each piece of information to be recommended and each set object to be recommended (for example, an information display position where the set object is clicked by the user to view when each set object to be recommended is displayed according to the display sequence, the number of the set objects of the exposed set object and the like are determined in a simulation mode). And the profit simulation is to simulate the profit of the set objects to be recommended, which is determined by simulating the browsing condition of the user on the information to be recommended and the set objects to be recommended, on the user and the service platform.
Specifically, the service platform may determine, according to the profit simulation, a recommendation effect characterization value corresponding to each to-be-recommended setting object, where the recommendation effect characterization value is used to characterize profits generated by each to-be-recommended setting object for itself and the service platform, and if the setting object is an advertisement, the recommendation effect characterization value may be specifically represented by, for example, CPM, GMV, Takerate, or the like.
S108: and determining an incentive value of an incentive function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the incentive value, wherein the information recommendation model is used for recommending a set object to a user.
In the process of training the information recommendation model, the service platform needs to refer to the behavior characteristics of the user, recommend the setting object to the user, and also needs to refer to the cost of the setting object and the income generated by the service platform, so that the setting object recommended by the information recommendation model can be effectively ensured to meet the personal preference of the user to a certain extent, and the setting object and the income of the service platform recommending the setting object can be ensured.
Based on this, in the embodiment of the present specification, the service platform may determine, according to the determined representation value of the recommendation effect, an incentive value of an incentive function corresponding to the information recommendation model, and train the information recommendation model according to the incentive value. The reward function can be referred to the following formula:
Figure BDA0002772587580000121
k1 to k6 are preset parameters, biasgmv、Kgmv、POWgmv、Kres、POWresIs a preset parameter;
Δ gmv is used to represent the difference value of the total volume of the setting object generated when different setting objects are recommended to the user and the setting objects are recommended to the user according to different display orders;
the delta fe is used for representing the difference value of commissions generated when different setting objects are recommended to the user and the setting objects are recommended to the user according to different display sequences;
the delta cpm is used for expressing a difference value of gains of the set objects obtained by the service platform when the set objects are recommended to the user and are recommended to the user according to different display sequences;
Δ res is a difference value indicating the number of setting targets each time a setting target is recommended;
Figure BDA0002772587580000131
the system is used for expressing the influence of the preference of the user on the set object income (such as the income of the advertisement caused by the preference of the user) of the set object (because the preference of different users is different, the number of the set objects accepted by some users is more, the income obtained by the service platform based on the set objects is higher, the number of the set objects accepted by some users is lower, and the income obtained by the service platform based on the set objects is lower);
further, in the above formula,
Figure BDA0002772587580000132
a first influence factor for characterizing the influence of a user's individual on the revenue of a set object of the service platform, and
Figure BDA0002772587580000133
Figure BDA0002772587580000134
for controlling setting of number of objects to be set to be personal to userAnd a second influence factor of the influence generated by the set object recommendation effect, wherein the set object recommendation effect can be measured in two aspects, one is the income which can be obtained by the service platform recommendation set object, and the other is the self cost of the set object. That is, by
Figure BDA0002772587580000135
The number of setting objects recommended to the user is limited to be within a proper range from the aspect of service, poor information browsing experience may be brought to the user due to the fact that the number of the setting objects is too high, and the cost of the setting objects and the income which can be obtained by recommending the setting objects by the service platform may be influenced due to the fact that the number of the setting objects is low.
The first and second influence factors are not unique in expression form, and may be expressed by other formulas, which are not described in detail herein.
In this embodiment of the present specification, the service platform may implement training of the information recommendation model by adjusting and optimizing the model parameters included in the information recommendation model and the weight function, with the reward value of the reward function being the maximum optimization goal. That is, after a plurality of rounds of iterative training, the reward value of the reward function is increased continuously and converged in a value range, and then the training process of the information recommendation model is completed.
Of course, in addition to training the information recommendation model with the maximum reward value as the optimization goal, the information recommendation model may be trained by adjusting model parameters included in the information recommendation model and the weight function with a preset reward value as the optimization goal, that is, in the process of multiple rounds of iterative training, the reward value needs to be continuously close to the preset reward value, and after multiple rounds of iterative training, the reward value floats back and forth around the preset reward value, and then the training of the information recommendation model can be determined to be completed.
It can be seen from the above process that in the process of model training, not only the behavior characteristic data of the user is considered, but also recommendation attribute data capable of expressing the cost and the profit of the set object to a certain extent is considered, so that the set object recommended to the user by the trained information recommendation model can not only meet the personal preference of the user, but also meet the cost and the profit of the set object recommended by the set object and the service platform, thereby not only optimizing the information sorting of the information recommendation list, avoiding the unreasonable sorting of the recommendation information, bringing good user experience to the information browsing of the user, and effectively ensuring the provider of the set object and the profit of the service platform.
After training of the information recommendation model is completed, the embodiment of the present specification may recommend recommendation information to a user through the information recommendation model, as shown in fig. 2.
Fig. 2 is a schematic flow chart of an information recommendation method provided in an embodiment of the present specification.
S200: and acquiring historical behavior data when a user browses information on an information recommendation list containing a set object.
S202: and determining behavior characteristic data corresponding to the user according to the historical behavior data.
S204: and inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommendation model to obtain each setting object to be recommended and a display sequence of each setting object to be recommended.
S206: and recommending the setting objects to be recommended to the user in the information to be recommended in a mixed arrangement mode according to the display sequence.
In this embodiment of the present specification, a service platform may obtain historical behavior data when a user historically includes an information recommendation list of setting objects, determine behavior feature data corresponding to the user according to the historical behavior data of the user, input the behavior feature data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommendation model, obtain a display order of each setting object to be recommended and each setting object to be recommended, and further recommend each setting object to be recommended and each information to be recommended to the user in a mixed manner according to the display order. The content of S200 to S204 is basically the same as the above model training stage, and is not described in detail here. The difference between S206 and the model training is that in the model training process, each setting object to be recommended is input into the information recommendation simulation system, and each setting object to be recommended is recommended to an actual user.
It should be emphasized that the weight parameters of the same user determined by the information recommendation model at different times are not completely the same, because the historical behavior data generated each time the user browses information is changed from the previous historical behavior data, the behavior feature data corresponding to the user determined by the historical behavior data of the user at different times are not completely the same, the behavior feature data at different times are input into a preset weight function, and the determined weight parameters corresponding to the user are changed.
Certainly, the service platform may also record the determined weight parameters of each user, so that when the user browses the information again, the service platform does not need to input the behavior feature data of the user into the preset weight function again, determines the weight parameters through the weight function corresponding to the user, but may call the previously determined weight parameters of the user, and then determines each setting object to be recommended output by the information recommendation model according to the called weight parameters of the user and the recommendation attribute data corresponding to each candidate setting object. Therefore, the calculation of the weight parameters of the users every time can be avoided, and network resources are saved.
In the training process of the information recommendation model in the embodiment of the present specification, the service platform may recommend each determined setting object to be recommended to the user through the information recommendation model, and then train the information recommendation model according to behavior data generated by the user for the setting object to be recommended, as shown in fig. 3.
Fig. 3 is a schematic flow chart of information recommendation model training and information recommendation provided in an embodiment of the present disclosure.
In the embodiment of the present specification, the information recommendation model is divided into the use of an online information recommendation model and the training of an offline information recommendation model, and the information recommendation model to be trained, the information recommendation simulation system, and the reward function constitute a training part of the offline information recommendation model. The service platform can determine the behavior characteristic data of the user A according to the historical behavior data of the user A when browsing the information recommendation list containing the setting objects, input the behavior characteristic data and the recommendation attribute data corresponding to each candidate setting object into the information recommendation model to be trained to obtain the display sequence of each setting object to be recommended and each setting object to be recommended, then inputting the setting objects to be recommended and the information to be recommended into an information recommendation simulation system according to the display sequence for simulation recommendation so as to determine the recommendation effect representation values corresponding to the setting objects to be recommended, according to the representation value of the recommendation effect, determining the reward value of the reward function corresponding to the information recommendation model to be trained, and according to the reward value, and training the information recommendation model to be trained, wherein the trained offline information recommendation model is used for updating the online information recommendation model.
Then, the service platform can recommend an information recommendation list containing a set object to the user through the updated information recommendation model, after the user browses the information recommendation list, the historical behavior data of the user will change, and the service platform can train the offline information recommendation model according to the changed historical behavior data of the user.
Based on the same idea, the present specification further provides a device for training an information recommendation model, as shown in fig. 4.
Fig. 4 is a schematic structural diagram of an information recommendation model training apparatus provided in an embodiment of this specification, which specifically includes:
an obtaining module 400, configured to obtain historical behavior data when each user browses information on an information recommendation list including a set object;
a determining module 402, configured to determine, for each user, behavior feature data of the user according to the historical behavior data;
a recommending module 404, configured to input the behavior feature data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained, so as to obtain each setting object to be recommended and a display order of each setting object to be recommended;
the simulation module 406 is configured to input the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation, so as to determine recommendation effect representation values corresponding to the setting objects to be recommended in the display sequence;
the training module 408 is configured to determine, according to the recommendation effect characterization value, an incentive value of an incentive function corresponding to the information recommendation model, and train the information recommendation model according to the incentive value, where the information recommendation model is used to perform set object recommendation for a user.
Optionally, the obtaining module 400 is specifically configured to obtain the historical behavior data includes: the number of the setting objects which are actually clicked and browsed by the user in the setting objects displayed by the information recommendation list, the number of the setting objects which are not clicked and browsed by the user in the setting objects displayed by the information recommendation list, and the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
Optionally, the determining module 402 is specifically configured to determine the churn rate of the set objects of the user for the information recommendation list according to the number of set objects actually clicked and browsed by the user among the set objects displayed by the information recommendation list and the number of set objects not clicked and browsed by the user among the set objects displayed by the information recommendation list, where the churn rate of the set objects of the user for the information recommendation list is greater if the number of set objects not clicked and browsed by the user among the set objects displayed by the information recommendation list is higher.
Optionally, the determining module 402 is specifically configured to determine the information browsing depth of the user according to an average list length of a displayed part of each information recommendation list when the user browses each information recommendation list.
Optionally, the recommending module 404 is specifically configured to input the behavior feature data of the user into a preset weight function, determine a weight parameter corresponding to the user, and determine each setting object to be recommended output by the information recommending model to be trained according to the weight parameter corresponding to the user and recommendation attribute data corresponding to each candidate setting object.
Optionally, the training module 408 is specifically configured to train the information recommendation model by adjusting model parameters included in the information recommendation model and the weight function with the reward value being maximum as an optimization goal.
Optionally, the training module 408 is specifically configured to determine a first influence factor and a second influence factor according to the characteristic value of the recommendation effect, where the first influence factor is used to characterize an influence of a user person on a profit of a set object, the second influence factor is used to control an influence of the number of the set objects recommended to the user on the user person and a recommendation effect of the set object, and determine a reward value of a reward function corresponding to the information recommendation model according to the characteristic value of the recommendation effect, the first influence factor, and the second influence factor.
Fig. 5 is a schematic structural diagram of an information recommendation device provided in an embodiment of this specification, which specifically includes:
the obtaining module 500 is configured to obtain historical behavior data when a user browses information on an information recommendation list including a set object;
a determining module 502, configured to determine, according to the historical behavior data, behavior feature data corresponding to the user;
a recommending module 504, configured to input the behavior feature data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommending model, so as to obtain each setting object to be recommended and a display order of each setting object to be recommended, where the information recommending model is obtained by training through a training method of the information recommending model;
and the pushing module 506 is configured to recommend the setting objects to be recommended to the user in the information to be recommended in a mixed manner according to the display sequence.
Optionally, the recommending module 504 is specifically configured to input the behavior feature data into a preset weight function, determine a weight function corresponding to the user, determine that weight parameters determined by different users through the weight function are not identical, and determine, according to the weight parameter corresponding to the user and recommendation attribute data corresponding to each candidate setting object, each setting object to be recommended and an arrangement order of the setting objects to be recommended, which are output by the information recommending model.
The present specification also provides a computer-readable storage medium storing a computer program, which can be used to execute the training method of the information recommendation model provided in fig. 1 or the information recommendation method provided in fig. 2.
The present specification also provides a schematic structural diagram of the electronic device shown in fig. 6. As shown in fig. 6, at the hardware level, the training device of the information recommendation model includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include hardware required by other services. The processor reads a corresponding computer program from the non-volatile memory into the memory and then runs the computer program to implement the training method of the information recommendation model described in fig. 1 or the information recommendation method described above. Of course, besides the software implementation, the present specification does not exclude other implementations, such as logic devices or a combination of software and hardware, and the like, that is, the execution subject of the following processing flow is not limited to each logic unit, and may be hardware or logic devices.
In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.
The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.
The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functions of the various elements may be implemented in the same one or more software and/or hardware implementations of the present description.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention 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.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
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.
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 tape 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, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the description 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.
This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims (13)

1. A training method of an information recommendation model is characterized by comprising the following steps:
acquiring historical behavior data of each user when browsing information of an information recommendation list containing a set object;
for each user, determining behavior characteristic data of the user according to the historical behavior data;
inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained to obtain each setting object to be recommended and a display sequence of each setting object to be recommended;
inputting the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation so as to determine recommendation effect representation values corresponding to the setting objects to be recommended in the display sequence;
and determining an incentive value of an incentive function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the incentive value, wherein the information recommendation model is used for recommending a set object to a user.
2. The method of claim 1, wherein the historical behavior data comprises: the number of the setting objects which are actually clicked and browsed by the user in the setting objects displayed by the information recommendation list, the number of the setting objects which are not clicked and browsed by the user in the setting objects displayed by the information recommendation list, and the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
3. The method of claim 2, wherein determining the behavior feature data of the user according to the historical behavior data specifically comprises:
and determining the loss rate of the set objects of the user aiming at the information recommendation list according to the number of the set objects actually clicked and browsed by the user aiming at the set objects displayed by the information recommendation list and the number of the set objects not clicked and browsed by the user aiming at the set objects displayed by the information recommendation list, wherein the loss rate of the set objects is larger if the number of the set objects not clicked and browsed by the user aiming at the set objects displayed by the information recommendation list is higher.
4. The method of claim 2, wherein determining the behavior feature data of the user according to the historical behavior data specifically comprises:
and determining the information browsing depth of the user according to the average list length of the displayed part of each information recommendation list when the user browses each information recommendation list.
5. The method of claim 1, wherein the information recommendation model comprises: a weighting function;
inputting the behavior feature data and recommendation attribute data corresponding to each candidate setting object into an information recommendation model to be trained to obtain each setting object to be recommended, and the method specifically comprises the following steps:
inputting the behavior characteristic data of the user into a preset weight function, and determining a weight parameter corresponding to the user;
and determining each set object to be recommended output by the information recommendation model to be trained according to the weight parameter corresponding to the user and the recommendation attribute data corresponding to each candidate set object.
6. The method of claim 5, wherein training the information recommendation model based on the reward value comprises:
and training the information recommendation model by adjusting the information recommendation model and model parameters contained in the weight function according to the maximum reward value as an optimization target.
7. The method according to claim 1, wherein determining, according to the recommendation effect characterization value, a reward value of a reward function corresponding to the information recommendation model specifically includes:
determining a first influence factor and a second influence factor according to the recommendation effect representation value, wherein the first influence factor is used for representing the influence of the user individuals on the profit of the set object, and the second influence factor is used for controlling the influence of the number of the set objects recommended to the user on the recommendation effect of the user individuals and the set object;
and determining the reward value of the reward function corresponding to the information recommendation model according to the recommendation effect representation value, the first influence factor and the second influence factor.
8. An information recommendation method, comprising:
acquiring historical behavior data when a user browses information on an information recommendation list containing a set object;
determining behavior characteristic data corresponding to the user according to the historical behavior data;
inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommendation model to obtain each setting object to be recommended and a display sequence of each setting object to be recommended, wherein the information recommendation model is obtained by training through the method of the claims 1-7;
and recommending the setting objects to be recommended to the user in the information to be recommended in a mixed arrangement mode according to the display sequence.
9. The method according to claim 8, wherein the behavior feature data and the recommendation attribute data corresponding to each candidate setting object are input into a pre-trained information recommendation model to obtain each setting object to be recommended and an arrangement order of the setting objects to be recommended, and specifically includes:
inputting the behavior characteristic data into a preset weight function, and determining the weight function corresponding to the user, wherein the weight parameters determined by different users through the weight function are not identical;
and determining each set object to be recommended and the arrangement sequence of each set object to be recommended, which are output by the information recommendation model, according to the weight parameters corresponding to the user and the recommendation attribute data corresponding to each candidate set object.
10. An apparatus for training an information recommendation model, comprising:
the acquisition module is used for historical behavior data of each user when the information recommendation list containing the set object is browsed;
the determining module is used for determining the behavior characteristic data of each user according to the historical behavior data;
the recommendation module is used for inputting the behavior characteristic data and recommendation attribute data corresponding to each candidate set object into an information recommendation model to be trained to obtain each set object to be recommended and a display sequence of each set object to be recommended;
the simulation module is used for inputting the setting objects to be recommended and the information to be recommended, which are arranged according to the display sequence, into a preset information recommendation simulation system for simulation recommendation so as to determine corresponding recommendation effect representation values of the setting objects to be recommended in the display sequence;
and the training module is used for determining the reward value of the reward function corresponding to the information recommendation model according to the recommendation effect representation value, and training the information recommendation model according to the reward value, wherein the information recommendation model is used for recommending a set object to a user.
11. An information recommendation apparatus, comprising:
the acquisition module is used for acquiring historical behavior data when a user browses information on an information recommendation list containing a set object;
the determining module is used for determining behavior characteristic data corresponding to the user according to the historical behavior data;
a recommending module, configured to input the behavior feature data and recommendation attribute data corresponding to each candidate setting object into a pre-trained information recommending model, so as to obtain each setting object to be recommended and a display order of each setting object to be recommended, where the information recommending model is obtained by training according to the method of any one of claims 1 to 7;
and the pushing module is used for recommending the setting objects to be recommended to the user in the information to be recommended in a mixed arrangement mode according to the display sequence.
12. A computer-readable storage medium, characterized in that the storage medium stores a computer program which, when executed by a processor, implements the method of any of the preceding claims 1 to 7 or 8 to 9.
13. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the method of any of claims 1 to 7 or 8 to 9.
CN202011254218.1A 2020-11-11 2020-11-11 Training method of information recommendation model, information recommendation method and device Pending CN112418920A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011254218.1A CN112418920A (en) 2020-11-11 2020-11-11 Training method of information recommendation model, information recommendation method and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011254218.1A CN112418920A (en) 2020-11-11 2020-11-11 Training method of information recommendation model, information recommendation method and device

Publications (1)

Publication Number Publication Date
CN112418920A true CN112418920A (en) 2021-02-26

Family

ID=74781521

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011254218.1A Pending CN112418920A (en) 2020-11-11 2020-11-11 Training method of information recommendation model, information recommendation method and device

Country Status (1)

Country Link
CN (1) CN112418920A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112948686A (en) * 2021-03-25 2021-06-11 支付宝(杭州)信息技术有限公司 Position recommendation processing method and device
CN113010563A (en) * 2021-03-16 2021-06-22 北京三快在线科技有限公司 Model training and information recommendation method and device
CN113010564A (en) * 2021-03-16 2021-06-22 北京三快在线科技有限公司 Model training and information recommendation method and device
CN113010809A (en) * 2021-03-11 2021-06-22 北京三快在线科技有限公司 Information recommendation method and device
CN113282819A (en) * 2021-04-22 2021-08-20 北京沃东天骏信息技术有限公司 Recommendation information display method, device, equipment and storage medium
CN113343085A (en) * 2021-05-27 2021-09-03 北京三快在线科技有限公司 Information recommendation method and device, storage medium and electronic equipment
CN113516511A (en) * 2021-07-13 2021-10-19 中国工商银行股份有限公司 Financial product purchase prediction method and device and electronic equipment
CN113641892A (en) * 2021-07-14 2021-11-12 北京三快在线科技有限公司 Information recommendation method and device
CN116205232A (en) * 2023-02-28 2023-06-02 之江实验室 Method, device, storage medium and equipment for determining target model

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113010809A (en) * 2021-03-11 2021-06-22 北京三快在线科技有限公司 Information recommendation method and device
CN113010564B (en) * 2021-03-16 2022-06-10 北京三快在线科技有限公司 Model training and information recommendation method and device
CN113010564A (en) * 2021-03-16 2021-06-22 北京三快在线科技有限公司 Model training and information recommendation method and device
CN113010563A (en) * 2021-03-16 2021-06-22 北京三快在线科技有限公司 Model training and information recommendation method and device
CN113010563B (en) * 2021-03-16 2022-02-01 北京三快在线科技有限公司 Model training and information recommendation method and device
CN112948686A (en) * 2021-03-25 2021-06-11 支付宝(杭州)信息技术有限公司 Position recommendation processing method and device
CN112948686B (en) * 2021-03-25 2023-06-16 支付宝(杭州)信息技术有限公司 Position recommendation processing method and device
CN113282819A (en) * 2021-04-22 2021-08-20 北京沃东天骏信息技术有限公司 Recommendation information display method, device, equipment and storage medium
CN113343085A (en) * 2021-05-27 2021-09-03 北京三快在线科技有限公司 Information recommendation method and device, storage medium and electronic equipment
CN113516511A (en) * 2021-07-13 2021-10-19 中国工商银行股份有限公司 Financial product purchase prediction method and device and electronic equipment
CN113641892A (en) * 2021-07-14 2021-11-12 北京三快在线科技有限公司 Information recommendation method and device
CN116205232A (en) * 2023-02-28 2023-06-02 之江实验室 Method, device, storage medium and equipment for determining target model
CN116205232B (en) * 2023-02-28 2023-09-01 之江实验室 Method, device, storage medium and equipment for determining target model

Similar Documents

Publication Publication Date Title
CN112418920A (en) Training method of information recommendation model, information recommendation method and device
CN108596645B (en) Information recommendation method, device and equipment
CN110503206A (en) A kind of prediction model update method, device, equipment and readable medium
TW202004618A (en) Product recommendation method and device
TW201939379A (en) Information conversion rate prediction method and apparatus, and information recommendation method and apparatus
CN113688313A (en) Training method of prediction model, information pushing method and device
CN113641894A (en) Information recommendation method and device
CN113010795B (en) User dynamic image generation method, system, storage medium and electronic device
KR20090017268A (en) Method for updating interest keyword of user and system for executing the method
CN111222931A (en) Product recommendation method and system
CN112966187A (en) Information recommendation method and device
CN115048577A (en) Model training method, device, equipment and storage medium
CN114882311A (en) Training set generation method and device
CN110738562B (en) Method, device and equipment for generating risk reminding information
CN111415210A (en) Information display method and device
CN111191132A (en) Information recommendation method and device and electronic equipment
KR20200070082A (en) Virtual assistant domain selection analysis
CN110889037A (en) Model training method and device
CN108139900B (en) Communicating information about updates of an application
CN111177562B (en) Recommendation ordering processing method and device for target object and server
CN112733014A (en) Recommendation method, device, equipment and storage medium
CN111639269A (en) Site recommendation method and device
CN112287208A (en) User portrait generation method and device, electronic equipment and storage medium
CN114676351A (en) Method, device and equipment for determining recommended position and storage medium
CN114116813A (en) Information recommendation method and recommendation device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination