CN113010563B

CN113010563B - Model training and information recommendation method and device

Info

Publication number: CN113010563B
Application number: CN202110280634.7A
Authority: CN
Inventors: 李爽; 谢乾龙; 林龙; 刘一飞; 王兴星; 王栋
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Network Technology Co ltd; Beijing Sankuai Online Technology Co Ltd
Priority date: 2021-03-16
Filing date: 2021-03-16
Publication date: 2022-02-01
Anticipated expiration: 2041-03-16
Also published as: CN113010563A

Abstract

The specification discloses a method and a device for model training and information recommendation, which are used for obtaining a historical service record of a user and obtaining first sample data according to the historical service record, wherein the first sample data comprises to-be-recommended sample information, first label information and second label information. And then, inputting the sample information to be recommended into a first recommendation model to be trained to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information, and finally, training the first recommendation model according to the determined first recommendation degree, the determined first label information, the determined second recommendation degree, the determined second label information, and the determined first combination degree. When information recommendation is required to be performed according to the second recommendation mode for the user, information recommendation can be performed on the user through the first recommendation model, and therefore information recommendation can be accurately performed on the user.

Description

Model training and information recommendation method and device

Technical Field

The specification relates to the technical field of machine learning, in particular to a method and a device for model training and information recommendation.

Background

With the continuous development of information technology, users can execute various services on line, various service platforms are generated accordingly, and in order to enable the users to execute the services more conveniently and quickly, the service platforms recommend information to the users according to the preferences of the users.

In practical applications, generally, when a user enters a service platform, the service platform may directly recommend some information to the user according to the preference of the user, and this recommendation manner is referred to as a first recommendation manner. For example, dishes at a merchant, or at a merchant's home, may be recommended directly to the user in the take-away platform. For another example, in a shopping platform, goods and the like may be directly recommended to a user, and in the first recommendation mode, the service platform may generally train a machine learning model to recommend information to the user.

Currently, after information recommendation is performed on a user by a service platform in a first recommendation manner, it may be determined that the user has performed a service for a certain piece of recommendation information, and then other pieces of recommendation information that may be needed by the user are recommended to the user. In the prior art, due to the fact that the number of training samples is insufficient, the service platform can directly use the machine learning model trained in the first recommendation mode to perform information recommendation, but the mode is relatively inaccurate.

Therefore, how to accurately recommend information to a user is an urgent problem to be solved.

Disclosure of Invention

The present specification provides a method and apparatus for model training and information recommendation, 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 method of model training, comprising:

acquiring a historical service record of a user;

obtaining first sample data according to the historical service record, wherein the first sample data comprises sample information to be recommended, first tag information and second tag information, the first tag information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after recommending the sample information to be recommended to the user according to a first recommending mode, the second tag information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after recommending the sample information to the user according to a second recommending mode, and the first recommending mode comprises: when the user does not execute other services, the user carries out information recommendation to the user, and the second recommendation mode comprises the following steps: after the user executes the service corresponding to the designated recommendation information, information recommendation is carried out on the user;

inputting the sample information to be recommended into a first recommendation model to be trained, and obtaining a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information;

and training the first recommendation model according to the first recommendation degree, the first label information, the second recommendation degree, the second label information and the first combination degree.

Optionally, the first recommendation model comprises: the system comprises a first recommendation submodel, a second recommendation submodel and a third recommendation submodel, wherein the first recommendation submodel is used for recommending information to a user according to the first recommendation mode, and the second recommendation submodel is used for recommending information to the user according to the second recommendation mode;

inputting the sample information to be recommended into a first recommendation model to be trained, to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information, specifically including:

inputting the sample information to be recommended into the first recommending submodel to obtain the first recommending degree, inputting the sample information to be recommended into the second recommending submodel to obtain the second recommending degree, and inputting the sample information to be recommended into the third recommending submodel to obtain the first combining degree.

Optionally, the first recommended sub-model, the second recommended sub-model, and the third recommended sub-model share a same coding layer;

inputting the sample information to be recommended into the first recommending sub-model to obtain the first recommendation degree, inputting the sample information to be recommended into the second recommending sub-model to obtain the second recommendation degree, and inputting the sample information to be recommended into the third recommending sub-model to obtain the first combination degree, specifically comprising:

inputting the sample information to be recommended into the coding layer to obtain a feature vector corresponding to the sample information to be recommended;

inputting the feature vector into a decision network contained in the first recommendation submodel to obtain the first recommendation degree, inputting the feature vector into a decision network contained in the second recommendation submodel to obtain the second recommendation degree, and inputting the feature vector into a decision network contained in the third recommendation submodel to obtain the first combination degree.

Optionally, training the first recommendation model according to the first recommendation degree, the first tag information, the second recommendation degree, the second tag information, and the first combination degree specifically includes:

determining an optimization strategy according with the to-be-recommended sample information according to the to-be-recommended sample information and the specified recommendation information;

and training the first recommendation model according to the optimization strategy, the minimization of the deviation between the first recommendation degree and the first label information and the minimization of the deviation between the second recommendation degree and the second label information as optimization targets.

Optionally, determining an optimization strategy that the to-be-recommended sample information conforms to according to the to-be-recommended sample information and the specified recommendation information specifically includes:

if it is determined that the pairing combination formed by the to-be-recommended sample information and the specified recommendation information does not appear in the historical service record, determining the category matching degree between the to-be-recommended sample information and the specified recommendation information, and taking the category matching degree and the first combination degree in a negative correlation relationship as an optimization strategy corresponding to the to-be-recommended sample information;

and if it is determined that the matching combination formed by the to-be-recommended sample information and the specified recommendation information appears in the historical service record, taking the highest first combination degree as an optimization target as an optimization strategy corresponding to the to-be-recommended sample information.

Optionally, the method further comprises:

determining at least one piece of other recommended sample information sent to the user after the user executes at least part of service corresponding to the to-be-recommended sample information from the historical service record;

for each piece of other recommended sample information, second sample data corresponding to the other recommended sample information, where the second sample data includes third tag information and fourth tag information corresponding to the other recommended sample information, the third tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information when recommending the other recommended sample information to the user according to the first recommendation manner, and the fourth tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information after recommending the to-be-recommended sample information to the user according to the second recommendation manner;

inputting the information of the sample to be recommended and the information of the other recommended samples into a second recommendation model to be trained to obtain a third recommendation degree corresponding to the first recommendation mode, a fourth recommendation degree corresponding to the second recommendation mode and a second combination degree between the information of the sample to be recommended and the information of the other recommended samples;

and training the second recommendation model according to the third recommendation degree, the third label information, the fourth recommendation degree, the fourth label information and the second combination degree.

The present specification provides a method for information recommendation, including:

if it is monitored that the user executes the service corresponding to the target recommendation information, determining other recommendation information except the target recommendation information as first candidate recommendation information;

for each piece of first candidate recommendation information, inputting the first candidate recommendation information into a pre-trained first recommendation model to obtain recommendation degrees of the first candidate recommendation information for the user, wherein the recommendation degrees of the first candidate recommendation information are used as recommendation degrees corresponding to the first candidate recommendation information, and the first recommendation model is obtained by training through a model training method;

and determining information to be recommended from the first candidate recommendation information according to the recommendation degree corresponding to the first candidate recommendation information, and recommending the information to be recommended to the user.

Optionally, the method further comprises:

if it is monitored that the user executes specified operation aiming at the information to be recommended, determining at least one piece of supplementary recommendation information according to the information to be recommended;

and displaying the information to be recommended and the at least one piece of supplementary recommendation information to the user in a preset page.

Optionally, determining at least one piece of supplemental recommendation information according to the information to be recommended specifically includes:

acquiring a plurality of second candidate recommendation information;

for each piece of second candidate recommendation information, inputting the information to be recommended and the second candidate recommendation information into a pre-trained second recommendation model to obtain a recommendation degree corresponding to the second candidate recommendation information;

and determining at least one piece of supplementary recommendation information according to the recommendation degrees corresponding to the plurality of second candidate recommendation information.

The present specification provides an apparatus for model training, comprising:

the record acquisition module is used for acquiring the historical service record of the user;

a sample obtaining module, configured to obtain first sample data according to the historical service record, where the first sample data includes to-be-recommended sample information, first tag information, and second tag information, the first tag information is used to indicate whether the user executes at least part of a service corresponding to the to-be-recommended sample information after recommending the to-be-recommended sample information to the user according to a first recommending manner, the second tag information is used to indicate whether the user executes at least part of a service corresponding to the to-be-recommended sample information after recommending the to-be-recommended sample information to the user according to a second recommending manner, and the first recommending manner includes: when the user does not execute other services, the user carries out information recommendation to the user, and the second recommendation mode comprises the following steps: after the user executes the service corresponding to the designated recommendation information, information recommendation is carried out on the user;

the input module is used for inputting the sample information to be recommended into a first recommendation model to be trained to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode and a first combination degree between the sample information to be recommended and the specified recommendation information;

and the first training module is used for training the first recommendation model according to the first recommendation degree, the first label information, the second recommendation degree, the second label information and the first combination degree.

This specification provides an apparatus for information recommendation, including:

the monitoring module is used for determining other recommendation information except the target recommendation information as first candidate recommendation information if the fact that the user executes the service corresponding to the target recommendation information is monitored;

the input module is used for inputting the first candidate recommendation information into a pre-trained first recommendation model aiming at each piece of first candidate recommendation information to obtain the recommendation degree of the first candidate recommendation information aiming at the user, and the recommendation degree is used as the recommendation degree corresponding to the first candidate recommendation information, and the first recommendation model is obtained by training through a model training method;

and the recommending module is used for determining information to be recommended from the first candidate recommending information according to the recommending degree corresponding to the first candidate recommending information and recommending the information to be recommended to the user.

The present specification provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements the above-described method of model training or information recommendation.

The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the above method of model training or information recommendation when executing the program.

The technical scheme adopted by the specification can achieve the following beneficial effects:

in the method and apparatus for model training or information recommendation provided in this specification, a historical service record of a user is obtained, and first sample data is obtained according to the historical service record, where the first sample data includes sample information to be recommended, first tag information and second tag information, the first tag information is used to indicate whether the user executes at least part of a service corresponding to the sample information to be recommended after recommending the sample information to be recommended to the user according to a first recommendation manner, the second tag information is used to indicate whether the user executes at least part of a service corresponding to the sample information to be recommended after recommending the sample information to be recommended to the user according to a second recommendation manner, and the first recommendation manner includes: when the user does not execute other services, information recommendation is carried out on the user, and the second recommendation mode comprises the following steps: and after the user executes the service corresponding to the specified recommendation information, recommending the information to the user. And then, inputting the sample information to be recommended into a first recommendation model to be trained to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information, and finally, training the first recommendation model according to the determined first recommendation degree, the determined first label information, the determined second recommendation degree, the determined second label information, and the determined first combination degree. When information recommendation is required to be performed according to the second recommendation mode for the user, information recommendation can be performed to the user through the first recommendation model.

It can be seen from the above method that if the service platform needs to recommend some information to the user after the user has performed the service, the service platform can determine how to recommend information to the user through the first recommendation model, when the first recommendation model is trained, the first recommendation model can be trained not only through the data related to the first recommendation mode, but also through the data related to the second recommendation mode and the combination degree between the sample information to be recommended in the second recommendation mode and the designated recommendation information of at least part of services which are executed by the user, so that, the first recommendation model is more suitable for use when information is recommended to a user by the second recommendation method, therefore, compared with the prior art in which the machine learning model in the first recommendation method is directly used, information can be more accurately recommended to the user.

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 flow chart of a method of model training in the present specification;

FIG. 2 is a schematic structural diagram of a first recommendation model provided in this specification;

fig. 3 is a schematic diagram of an interface for recommending information to a user by a service platform provided in this specification;

FIG. 4 is a flow chart illustrating a method for information recommendation in the present specification;

FIG. 5 is a schematic diagram of an apparatus for model training provided herein;

FIG. 6 is a schematic diagram of an apparatus for information recommendation provided herein;

fig. 7 is a schematic diagram of an electronic device corresponding to fig. 1 or fig. 4 provided in the present 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.

Fig. 1 is a schematic flow chart of a model training method in this specification, which specifically includes the following steps:

s101: and acquiring the historical service record of the user.

S102: obtaining first sample data according to the historical service record, wherein the first sample data comprises sample information to be recommended, first tag information and second tag information, the first tag information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after recommending the sample information to be recommended to the user according to a first recommending mode, the second tag information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after recommending the sample information to the user according to a second recommending mode, and the first recommending mode comprises: when the user does not execute other services, the user carries out information recommendation to the user, and the second recommendation mode comprises the following steps: and after the user executes the service corresponding to the specified recommendation information, recommending the information to the user.

In practical application, the service platform needs to recommend information to a user through multiple recommendation modes, the first recommendation mode is that the service platform can directly recommend information to the user when the user does not execute any service, for example, when the user just enters a home page of the takeout platform through a terminal, the takeout platform can recommend some takeout information to the user, such as dishes, fruits, drinks and the like. The second recommendation mode is that the service platform can recommend information to the user after the user executes the service corresponding to the specified recommendation information. Still taking the example of the takeout service, after the user orders the takeout of a certain dish, the user may be recommended to take out other dishes besides the certain dish.

In the information recommendation service, the first recommendation mode is widely applied, so that a large number of training samples and mature machine learning models exist in the first recommendation mode, a service platform can accurately recommend information to a user according to the first recommendation mode, the second recommendation mode is not widely applied compared with the first recommendation mode, the service platform can train the first recommendation mode through the method provided by the specification, and therefore information can be accurately recommended to the user when information is recommended through the second recommendation mode.

Based on this, the service platform can obtain the historical service record of the user, and obtain the first sample data according to the historical service record, wherein the first sample data comprises the sample information to be recommended, the first label information and the second label information. The first label information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after the sample information to be recommended is recommended to the user according to the first recommendation mode, and the second label information is used for indicating whether the user executes at least part of service corresponding to the sample information to be recommended after the sample information to be recommended is recommended to the user according to the second recommendation mode. The first recommendation method may refer to that the user performs information recommendation to the user when the user does not execute other services, and the second recommendation method may refer to that the user performs information recommendation to the user after the user executes the service corresponding to the specified recommendation information.

The to-be-recommended sample information mentioned herein may refer to recommendation information historically shown to a user, and the to-be-recommended sample information may further include feature information of the recommendation information, for example, a category corresponding to the to-be-recommended sample information, a price corresponding to the to-be-recommended sample information, a gender and an age of the user, recommendation information that is clicked, placed an order, and the like.

The first tag information mentioned above is related to the first recommendation manner, still taking the takeaway service as an example, the first tag information may be whether the user clicks or orders the takeaway information after directly recommending a certain takeaway information to the user, that is, the at least part of the service may include at least part of the operation in the takeaway service corresponding to the takeaway information, for example, clicking, ordering, commenting, and the like on the takeaway information. Correspondingly, the second tag information is related to the second recommendation mode, still taking the takeout service as an example, the second tag information may refer to that the user recommends other takeout information to the user after purchasing a single takeout, and whether the user clicks or places an order for the other takeout information. The above-mentioned specified recommendation information may refer to the information that the user takes out the purchase of the purchase completion.

S103: and inputting the sample information to be recommended into a first recommendation model to be trained to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information.

After the service platform acquires the sample information to be recommended, the service platform may input the sample information to be recommended into a first recommendation model to be trained, to obtain a first recommendation degree corresponding to a first recommendation mode, a second recommendation degree corresponding to a second recommendation mode, and a first combination degree between the sample information to be recommended and specified recommendation information, where the first recommendation degree corresponding to the first recommendation mode mentioned here is a recommendation degree predicted for the user, and the recommendation degree may be a predicted click rate, a rejection rate, or the like. The first combination degree between the to-be-recommended sample information and the specified recommendation information may refer to a degree that recommendation information corresponding to the to-be-recommended sample information and the specified recommendation information can be matched for recommendation, that is, the first combination degree may also represent a recommendation degree that recommendation information corresponding to the to-be-recommended sample information can be recommended to a user after the user executes a service corresponding to the specified recommendation information.

In this specification, the first recommendation model includes: the recommendation system comprises a first recommendation submodel, a second recommendation submodel and a third recommendation submodel, wherein the first recommendation submodel is used for recommending information to a user according to a first recommendation mode, the second recommendation submodel is used for recommending information to the user according to a second recommendation mode, and the third recommendation submodel is used for recommending information to the user through a predicted second degree of combination.

Therefore, the service platform can input the sample information to be recommended into the first recommending submodel to obtain a first recommendation degree, input the sample information to be recommended into the second recommending submodel to obtain a second recommendation degree, and input the sample information to be recommended into the third recommending submodel to obtain a first combination degree.

In practical applications, the first recommended sub-model, the second recommended sub-model, and the third recommended sub-model may share the same coding layer, as shown in fig. 2.

Fig. 2 is a schematic structural diagram of a first recommendation model provided in this specification.

As can be seen from fig. 2, the service platform may input the sample information to be recommended into the coding layer to obtain a feature vector corresponding to the sample information to be recommended, then input the feature vector into a decision network included in a first recommendation submodel to obtain a first recommendation degree, input the feature vector into a decision network included in a second recommendation submodel to obtain a second recommendation degree, and input the feature vector into a decision network included in a third recommendation submodel to obtain a first combination degree.

S104: and training the first recommendation model according to the first recommendation degree, the first label information, the second recommendation degree, the second label information and the first combination degree.

In this specification, the service platform obtains a first recommendation degree, a second recommendation degree, and a first combination degree through a first recommendation model, and may train the first recommendation model according to the first recommendation degree, the second recommendation degree, the first tag information, the second tag information, and the first combination degree.

Specifically, the service platform may determine an optimization strategy that the to-be-recommended sample information conforms to according to the to-be-recommended sample information and the specified recommendation information, and train the first recommendation model according to the optimization strategy, minimizing a deviation between the first recommendation degree and the first tag information, and minimizing a deviation between the second recommendation degree and the second tag information as optimization targets.

That is to say, the optimization strategy may be determined according to a relationship between the to-be-recommended sample information and the specified recommendation information, and specifically, if the service platform determines that a matching combination of the to-be-recommended sample information and the specified recommendation information appears in the historical service record, the highest first combination degree may be used as an optimization target, and the optimization strategy is used as an optimization strategy corresponding to the to-be-recommended sample information. The first combination degree can be determined through the similarity between the feature vector of the recommendation information corresponding to the sample information to be recommended and the feature vector of the specified recommendation information.

The occurrence of the combination of the to-be-recommended sample information and the specified recommendation information in the historical service record may be that the user has historically performed the service corresponding to the recommendation information corresponding to the to-be-recommended sample information and the specified recommendation information within a set time. Or, the user executes the service of the category to which the recommendation information belongs and the category to which the specified recommendation information belongs within the set time historically.

For example, if the category to which the recommendation information corresponding to the sample information to be recommended belongs is a staple food type, the category to which the specified recommendation information belongs is a beverage type, and the user orders takeout of the staple food type and the beverage type within 30 minutes in the history, it can be determined that a pairing combination of the sample information to be recommended and the specified recommendation information appears in the history service record.

Therefore, the optimization strategy determined by the to-be-recommended sample information and the specified recommendation information is that when the service platform determines that a matching combination formed by the to-be-recommended sample information and the specified recommendation information appears in the historical service record, a first matching degree corresponding to the to-be-recommended sample information and the specified recommendation information should be the highest, and since a value of the first matching degree is the highest 1, that is, it is desirable that the first matching degree between the to-be-recommended sample information and the specified recommendation information approaches to 1.

The optimization strategy can be represented by an optimization objective related to the first degree of combination included in the optimization objective corresponding to the first prediction model, and the optimization objective related to the first degree of combination can be specifically an optimization objective of minimizing a deviation between the first degree of combination and the second labeled information. The first recommendation model obtained through the training of the optimization strategy can enable the predicted second recommendation degree to be in positive correlation with the first recommendation degree, namely, when a matching combination formed by the sample information to be recommended and the specified recommendation information appears in the historical service record, the second recommendation degree predicted by the first recommendation model is high.

If the service platform determines that the pairing combination formed by the to-be-recommended sample information and the specified recommendation information does not appear in the historical service record, the class matching degree between the to-be-recommended sample information and the specified recommendation information can be determined, and the class matching degree and the first combination degree are in a negative correlation relationship and serve as an optimization strategy corresponding to the to-be-recommended sample information.

The category matching degree mentioned here may represent a degree of similarity between the sample information to be recommended and the designated recommendation information, where the closer the category between the sample information to be recommended and the designated recommendation information is, the higher the category matching degree is, and the less the category between the sample information to be recommended and the designated recommendation information is, the lower the category matching degree is. The optimization strategy aims to enable the predicted second recommendation degree of the first recommendation model after training to be lower when the type of the sample information to be recommended is similar to that of the specified recommendation information.

The optimization strategy can be obtained by negating a loss function corresponding to the optimization target related to the first combination degree when the combination of the to-be-recommended sample information and the specified recommendation information appears in the historical service record, that is, the optimization strategy has an opposite effect to the optimization target when the combination of the to-be-recommended sample information and the specified recommendation information appears in the historical service record.

In practical application, when the service platform recommends information for the user through the second recommendation method, a specific form may be as shown in fig. 3.

Fig. 3 is a schematic interface diagram of a service platform for recommending information to a user provided in this specification.

As can be seen from fig. 3, after the user executes the service for specifying the recommendation information in the first page, the service platform recommends some other recommendation information for the user below the first page, where the recommendation information may be recommendation information corresponding to the sample information to be recommended, and after the user clicks some recommendation information in the first page, the terminal displays a second page in which information recommendation is performed based on the recommendation information clicked by the user in the first page. The first page can use the first recommendation model to recommend information, and the second page can use the second recommendation model to recommend information. This second recommendation model also needs to be trained in a similar way as the first recommendation model.

Specifically, the service platform may determine, from the historical service record, at least one piece of other recommended sample information that is sent to the user after the user executes at least part of the service corresponding to the to-be-recommended sample information, and for each piece of other recommended sample information, second sample data corresponding to the other recommended sample information, where the second sample data includes third tag information and fourth tag information corresponding to the other recommended sample information, the third tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information when the user recommends the other recommended sample information according to the first recommendation method, and the fourth tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information after the user recommends the to-be-recommended sample information according to the second recommendation method.

The service platform inputs the sample information to be recommended and the other recommended sample information into a second recommended model to be trained, so that a third recommendation degree corresponding to the first recommended mode, a fourth recommendation degree corresponding to the second recommended mode and a second combination degree between the sample information to be recommended and the other recommended sample information can be obtained, then the service platform can train the second recommended model according to the third recommendation degree, the third label information, the fourth recommendation degree, the fourth label information and the second combination degree, the specific training mode is similar to that of the first predicted model, and the difference is that the service platform can directly train the second predicted model by taking the highest second combination degree between the sample information to be recommended and the other recommended sample information as an optimization target.

The above is all the explanation of the method from the perspective of model training, and the following is the explanation from the perspective of actually recommending information on a service platform.

Fig. 4 is a schematic flowchart of an information recommendation method provided in this specification, which specifically includes the following steps:

s401: and if the situation that the user executes the service corresponding to the target recommendation information is monitored, determining other recommendation information except the target recommendation information as first candidate recommendation information.

S401: and inputting the first candidate recommendation information into a pre-trained first recommendation model aiming at each piece of first candidate recommendation information to obtain the recommendation degree of the first candidate recommendation information aiming at the user, wherein the recommendation degree is used as the recommendation degree corresponding to the first candidate recommendation information, and the first recommendation model is obtained by training through the model training method.

S401: and determining information to be recommended from the first candidate recommendation information according to the recommendation degree corresponding to the first candidate recommendation information, and recommending the information to be recommended to the user.

In fig. 4, the service platform performs information recommendation on the user according to the second recommendation manner, and specifically, after monitoring that the user executes a service corresponding to the target recommendation information, the service platform may determine, as each first candidate recommendation information, each piece of recommendation information other than the target recommendation information. The target recommendation information mentioned here is similar to the above-mentioned specified recommendation information, that is, the above-mentioned specified recommendation information corresponds to sample information to be recommended, and the target recommendation information corresponds to the first candidate recommendation information, both of which are recommendation information for the user to perform the service execution.

Then, for each piece of first candidate recommendation information, inputting the first candidate recommendation information into a pre-trained first recommendation model to obtain a recommendation degree of the first candidate recommendation information for a user, and taking the recommendation degree as a recommendation degree corresponding to the first candidate recommendation information, wherein the first recommendation model is obtained by training in the model training manner, and the service platform can determine information to be recommended from each piece of first candidate recommendation information according to the recommendation degree corresponding to each piece of first candidate recommendation information and recommend the information to be recommended to the user.

After recommending information to be recommended to a user, the service platform may determine at least one piece of supplemental recommendation information according to the information to be recommended after monitoring that the user performs a specified operation on a certain piece of information to be recommended, and display the information to be recommended and the at least one piece of supplemental recommendation information in a preset page to the user, where the specified operation mentioned herein may be specifically set, for example, the specified operation may be a click operation on a link corresponding to the information to be recommended. The preset page referred to herein may refer to the above-mentioned second page.

The service platform can acquire a plurality of second candidate recommendation information after monitoring that a user performs specified operation on information to be recommended, inputs the information to be recommended and the second candidate recommendation information into a pre-trained second recommendation model for each second candidate recommendation information, obtains recommendation degrees corresponding to the second candidate recommendation information, and determines at least one piece of supplementary recommendation information according to the recommendation degrees corresponding to the plurality of second candidate recommendation information. The second candidate recommendation information mentioned here may be determined from the first candidate recommendation information or each piece of information to be recommended, and may of course be candidate recommendation information determined separately.

The method can be seen in that a service platform can train a first recommendation model and a second recommendation model through the method, and through the first recommendation model, the service platform can recommend information to a user after the user executes a service aiming at target recommendation information, that is, recommend information to the user according to a second recommendation mode.

The first recommendation model further includes a third recommendation submodel for predicting a combination degree between the target recommendation information and the first candidate recommendation information, and the third recommendation submodel enables a second recommendation degree predicted by the first recommendation model to be considered in relation to the target recommendation information and the first candidate recommendation information, that is, if a pairing combination of the target recommendation information and the first candidate recommendation information appears in the user history service record, the second recommendation degree is high, and if categories between the target recommendation information and the first candidate recommendation information are too close, the second recommendation degree is low.

Based on the same idea, the present specification further provides a corresponding apparatus for model training and information recommendation, as shown in fig. 5 or fig. 6.

Fig. 5 is a schematic diagram of a model training apparatus provided in this specification, which specifically includes:

a record obtaining module 501, configured to obtain a historical service record of a user;

a sample obtaining module 502, configured to obtain first sample data according to the historical service record, where the first sample data includes to-be-recommended sample information, first tag information, and second tag information, the first tag information is used to indicate whether the user executes at least part of a service corresponding to the to-be-recommended sample information after recommending the to-be-recommended sample information to the user according to a first recommending manner, the second tag information is used to indicate whether the user executes at least part of a service corresponding to the to-be-recommended sample information after recommending the to-be-recommended sample information to the user according to a second recommending manner, and the first recommending manner includes: when the user does not execute other services, the user carries out information recommendation to the user, and the second recommendation mode comprises the following steps: after the user executes the service corresponding to the designated recommendation information, information recommendation is carried out on the user;

the input module 503 is configured to input the sample information to be recommended into a first recommendation model to be trained, so as to obtain a first recommendation degree corresponding to the first recommendation mode, a second recommendation degree corresponding to the second recommendation mode, and a first combination degree between the sample information to be recommended and the specified recommendation information;

a first training module 504, configured to train the first recommendation model according to the first recommendation degree, the first tag information, the second recommendation degree, the second tag information, and the first combination degree.

the input module 503 is specifically configured to input the sample information to be recommended into the first recommending sub-model to obtain the first recommendation degree, input the sample information to be recommended into the second recommending sub-model to obtain the second recommendation degree, and input the sample information to be recommended into the third recommending sub-model to obtain the first combination degree.

the input module 503 is specifically configured to input the to-be-recommended sample information into the coding layer, so as to obtain a feature vector corresponding to the to-be-recommended sample information; inputting the feature vector into a decision network contained in the first recommendation submodel to obtain the first recommendation degree, inputting the feature vector into a decision network contained in the second recommendation submodel to obtain the second recommendation degree, and inputting the feature vector into a decision network contained in the third recommendation submodel to obtain the first combination degree.

Optionally, the first training module 504 is specifically configured to determine, according to the to-be-recommended sample information and the specified recommendation information, an optimization strategy that the to-be-recommended sample information conforms to; and training the first recommendation model according to the optimization strategy, the minimization of the deviation between the first recommendation degree and the first label information and the minimization of the deviation between the second recommendation degree and the second label information as optimization targets.

Optionally, the first training module 504 is specifically configured to, if it is determined that the pairing combination formed by the to-be-recommended sample information and the specified recommendation information does not appear in the historical service record, determine a class matching degree between the to-be-recommended sample information and the specified recommendation information, and use the class matching degree and the first combining degree in a negative correlation as an optimization strategy corresponding to the to-be-recommended sample information; and if it is determined that the matching combination formed by the to-be-recommended sample information and the specified recommendation information appears in the historical service record, taking the highest first combination degree as an optimization target as an optimization strategy corresponding to the to-be-recommended sample information.

Optionally, the apparatus further comprises:

a second training module 505, configured to determine, from the historical service record, at least one piece of other recommended sample information that is sent to the user after the user executes at least part of the service corresponding to the to-be-recommended sample information; for each piece of other recommended sample information, second sample data corresponding to the other recommended sample information, where the second sample data includes third tag information and fourth tag information corresponding to the other recommended sample information, the third tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information when recommending the other recommended sample information to the user according to the first recommendation manner, and the fourth tag information is used to indicate whether the user executes at least part of the service corresponding to the other recommended sample information after recommending the to-be-recommended sample information to the user according to the second recommendation manner; inputting the information of the sample to be recommended and the information of the other recommended samples into a second recommendation model to be trained to obtain a third recommendation degree corresponding to the first recommendation mode, a fourth recommendation degree corresponding to the second recommendation mode and a second combination degree between the information of the sample to be recommended and the information of the other recommended samples; and training the second recommendation model according to the third recommendation degree, the third label information, the fourth recommendation degree, the fourth label information and the second combination degree.

Fig. 6 is a schematic diagram of an information recommendation apparatus provided in this specification, which specifically includes:

the monitoring module 601 is configured to determine, if it is monitored that a user has executed a service corresponding to target recommendation information, each piece of other recommendation information except the target recommendation information as each piece of first candidate recommendation information;

an input module 602, configured to input, for each piece of first candidate recommendation information, the piece of first candidate recommendation information into a pre-trained first recommendation model, to obtain a recommendation degree of the piece of first candidate recommendation information for the user, where the recommendation degree is used as a recommendation degree corresponding to the piece of first candidate recommendation information, and the first recommendation model is obtained by training through a model training method;

and the recommending module 603 is configured to determine information to be recommended from each piece of first candidate recommendation information according to the recommendation degree corresponding to each piece of first candidate recommendation information, and recommend the information to be recommended to the user.

Optionally, the apparatus further comprises:

a display module 604, configured to determine at least one piece of supplemental recommendation information according to the information to be recommended if it is monitored that the user performs a specified operation on the information to be recommended; and displaying the information to be recommended and the at least one piece of supplementary recommendation information to the user in a preset page.

Optionally, the presentation module 604 is specifically configured to obtain a plurality of second candidate recommendation information; for each piece of second candidate recommendation information, inputting the information to be recommended and the second candidate recommendation information into a pre-trained second recommendation model to obtain a recommendation degree corresponding to the second candidate recommendation information; and determining at least one piece of supplementary recommendation information according to the recommendation degrees corresponding to the plurality of second candidate recommendation information.

The present specification also provides a computer-readable storage medium storing a computer program, which can be used to execute the method for model training and information recommendation shown in fig. 1 or fig. 4.

This specification also provides a schematic block diagram of the electronic device shown in fig. 7. As shown in fig. 7, at the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, but may also include hardware required for 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 model training and information recommendation method described in fig. 1 or fig. 4. 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

1. A method of model training, comprising:

acquiring a historical service record of a user;

2. The method of claim 1, wherein the first recommendation model comprises: the system comprises a first recommendation submodel, a second recommendation submodel and a third recommendation submodel, wherein the first recommendation submodel is used for recommending information to a user according to the first recommendation mode, and the second recommendation submodel is used for recommending information to the user according to the second recommendation mode;

3. The method of claim 2, wherein the first recommended sub-model, the second recommended sub-model, and the third recommended sub-model share a same encoding layer;

4. The method of claim 1, wherein training the first recommendation model according to the first recommendation, the first label information, the second recommendation, the second label information, and the first combination degree comprises:

5. The method according to claim 4, wherein determining an optimization strategy that the to-be-recommended sample information conforms to according to the to-be-recommended sample information and the specified recommendation information specifically includes:

6. The method of claim 1, wherein the method further comprises:

7. A method for information recommendation, comprising:

for each piece of first candidate recommendation information, inputting the first candidate recommendation information into a pre-trained first recommendation model to obtain a recommendation degree of the first candidate recommendation information for the user, wherein the recommendation degree of the first candidate recommendation information is used as a recommendation degree corresponding to the first candidate recommendation information, and the first recommendation model is obtained by training according to the method of any one of claims 1 to 6;

8. The method of claim 7, wherein the method further comprises:

if it is monitored that the user executes specified operation on the information to be recommended, determining at least one piece of supplementary recommendation information according to the information to be recommended;

9. The method according to claim 8, wherein determining at least one piece of supplemental recommendation information according to the information to be recommended specifically comprises:

acquiring a plurality of second candidate recommendation information;

10. An apparatus for model training, comprising:

11. An apparatus for information recommendation, comprising:

an input module, configured to input, for each piece of first candidate recommendation information, the piece of first candidate recommendation information into a pre-trained first recommendation model, to obtain a recommendation degree of the piece of first candidate recommendation information for the user, where the recommendation degree is used as a recommendation degree corresponding to the piece of first candidate recommendation information, and the first recommendation model is obtained by training according to the method of any one of claims 1 to 6;

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 claims 1 to 6 or 7 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 6 or 7 to 9.