CN113887596A

CN113887596A - Model training and information pushing method and device

Info

Publication number: CN113887596A
Application number: CN202111120081.5A
Authority: CN
Inventors: 钱心远; 李翔
Original assignee: Beijing Sankuai Online Technology Co Ltd
Current assignee: Beijing Sankuai Online Technology Co Ltd
Priority date: 2021-09-24
Filing date: 2021-09-24
Publication date: 2022-01-04

Abstract

The specification discloses a model training and information pushing method and device, which can determine gains of unit resource quantity relative to commodities through a pre-trained gain estimation model based on commodity characteristics of the commodities. And then, determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and pushing the information to the user. And estimating the gain of the unit resource quantity relative to each commodity through a gain estimation model, and determining the push information corresponding to each commodity pushed to the user according to the gain of the unit resource quantity relative to each commodity. The effectiveness of information pushing is improved, the resource amount corresponding to the pushed information is more in line with the user requirements, and the waste of the resources of the pushed information is reduced.

Description

Model training and information pushing method and device

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a model training and information pushing method and device.

Background

With the development of internet technology, more and more users execute services through a network platform, such as purchasing goods through a merchant platform, communicating through a social platform, and the like.

In order to improve the retention rate of the user, the network platforms enable the user to obtain corresponding services according to resources corresponding to the pushed information in a mode of pushing the information to the user, and user experience is improved.

Currently, when information is pushed, each network platform generally pushes different types of information to each user at random, so that each user obtains a corresponding service according to a resource corresponding to the obtained pushed information.

However, since the services required by each user are different, the allocated resources of the push information often do not meet the actual needs of the user, and the resources corresponding to the push information are wasted.

Disclosure of Invention

The embodiment of the specification provides a model training and information pushing method and device, which are used for partially solving the problems in the prior art.

The embodiment of the specification adopts the following technical scheme:

the model training method provided by the specification comprises the following steps:

for each commodity, acquiring commodity characteristics and historical data of the commodity, wherein the historical data at least comprises a first operated frequency and a second operated frequency of the commodity in history; the first operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is not pushed to the user; the second operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is pushed to the user;

determining the gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the total amount of resources contained in pushing information corresponding to the commodity historically pushed to a user;

taking the commodity characteristics of each commodity as each training sample, and taking the gain of the unit resource amount relative to each commodity as the label of each training sample;

aiming at each training sample, inputting the training sample into a gain estimation model to be trained, determining the gain to be optimized of the unit resource quantity output by the gain estimation model relative to the commodity, and adjusting model parameters in the gain estimation model to be trained by taking the difference between the gain to be optimized and the label of the training sample as a target;

the gain estimation model is used for estimating gains of the unit resource quantity relative to the commodities, and determining the information to be pushed corresponding to the commodities and pushing the information based on the gains of the unit resource quantity relative to the commodities and the total resource quantity contained in the information to be pushed corresponding to the commodities at present.

Optionally, determining a gain of the unit resource amount relative to the commodity according to the historical first operated frequency and second operated frequency of the commodity and a total resource amount included in pushing information corresponding to the commodity historically pushed to the user, specifically including:

determining the total amount of resources actually used by the user in history from the total amount of resources contained in the pushing information corresponding to the commodity historically pushed to the user;

and determining the gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the historical total amount of resources actually used by the user.

Optionally, the first operated times include a first access amount and a first order amount of the user to the product when the push information corresponding to the product is not pushed to the user, and the second operated times include a second access amount and a second order amount of the user to the product when the push information corresponding to the product is pushed to the user;

determining a gain of a unit resource amount relative to the commodity according to the historical first operated frequency and second operated frequency of the commodity and a total resource amount contained in pushing information corresponding to the commodity historically pushed to a user, specifically comprising:

determining a first ordering conversion rate of the commodity according to a first historical visit amount and a first ordering amount of the commodity;

determining a second ordering conversion rate of the commodity according to a second historical visit amount and a second ordering amount of the commodity;

and determining the gain of the unit resource amount relative to the commodity according to the first ordering conversion rate, the second visit amount and the total amount of resources contained in push information corresponding to the commodity historically pushed to the user.

The information push method provided by the present specification includes:

acquiring commodity characteristics of each commodity;

inputting the commodity characteristics of each commodity into a pre-trained gain estimation model, and determining the gain of unit resource quantity relative to the commodity; the gain estimation model is obtained by training by adopting the model training method;

determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and pushing the information to be pushed corresponding to each commodity to a user according to the type of the information to be pushed corresponding to each commodity;

the resource amount corresponding to the different types of push information is different.

Optionally, determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount included in the information to be pushed corresponding to each commodity, specifically including:

dividing a plurality of commodity groups according to the sequencing of the unit resource amount relative to the gain brought by each commodity;

determining the type of the information to be pushed corresponding to each commodity group according to the quantity of the commodities in each commodity group and the total amount of resources contained in the information to be pushed corresponding to each commodity;

determining the type of the information to be pushed corresponding to each commodity according to the type of the information to be pushed corresponding to each commodity group; the types of the information to be pushed corresponding to the commodities in the same commodity group are the same.

Optionally, the method further comprises:

for each user, determining the ordering probability of the user under the condition that the user does not receive the push information corresponding to the commodity and the ordering probability under the condition that the user receives the push information corresponding to the commodity according to the portrait information and the historical behavior information of the user;

and pushing the push information corresponding to the commodity to the user according to the ordering probability of the user under the condition that the push information corresponding to the commodity is not received, the ordering probability of the user under the condition that the push information corresponding to the commodity is received and the type of the push information corresponding to the commodity.

This specification provides a model training device, comprising:

the acquisition module is configured to acquire commodity characteristics and historical data of each commodity, wherein the historical data at least comprises a first operated frequency and a second operated frequency of the commodity in history; the first operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is not pushed to the user; the second operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is pushed to the user;

the determining module is configured to determine a gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the total amount of resources contained in pushing information corresponding to the commodity historically pushed to a user;

the sample preparation module is configured to take the commodity characteristics of each commodity as each training sample, and take the gain of the unit resource amount relative to each commodity as the label of each training sample;

the training module is configured to input the training sample into a gain estimation model to be trained for each training sample, determine gain to be optimized, which is output by the gain estimation model and is caused by the unit resource amount relative to the commodity, and adjust model parameters in the gain estimation model to be trained by taking the difference between the gain to be optimized and the label of the training sample as a target, wherein the gain estimation model is used for estimating the gain caused by the unit resource amount relative to each commodity, and determining information to be pushed corresponding to each commodity and pushing the information based on the gain caused by the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each current commodity.

This specification provides an information push apparatus, including:

the acquisition module is configured to acquire the commodity characteristics of each commodity;

the determining module is configured to input the commodity characteristics of each commodity into a pre-trained gain estimation model and determine the gain of the unit resource quantity relative to the commodity; the gain estimation model is obtained by training by adopting the model training method;

the distribution module is configured to determine the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and push the information to be pushed corresponding to each commodity to a user according to the type of the information to be pushed corresponding to each commodity, wherein the resource amounts corresponding to different types of push information are different.

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

The embodiment of the specification adopts at least one technical scheme which can achieve the following beneficial effects:

in the present specification, the gain of the unit resource amount for each commodity can be determined by a gain estimation model trained in advance based on the commodity characteristics of each commodity. And then, determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and pushing the information to the user. And estimating the gain of the unit resource quantity relative to each commodity through a gain estimation model, and determining the push information corresponding to each commodity pushed to the user according to the gain of the unit resource quantity relative to each commodity. The effectiveness of information pushing is improved, the resource amount corresponding to the pushed information is more in line with the user requirements, and the waste of the resources of the pushed information is reduced.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

fig. 1 is a schematic flow chart of a model training method provided in an embodiment of the present disclosure;

fig. 2 is a schematic flowchart of an information pushing method provided in an embodiment of the present disclosure;

FIG. 3 is a schematic structural diagram of a model training apparatus according to an embodiment of the present disclosure;

fig. 4 is a schematic structural diagram of an information pushing apparatus provided in an embodiment of the present disclosure;

fig. 5 is a schematic view of an electronic device implementing a model training method or an information pushing method according to an embodiment of the present disclosure.

Detailed Description

In order to make the objects, technical solutions and advantages of the present disclosure more apparent, 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 should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments obtained by a person skilled in the art without making any inventive step based on the embodiments in the description belong to the protection scope of the present application.

The present specification provides a model training method, and the following describes technical solutions provided in embodiments of the present application in detail with reference to the accompanying drawings.

Fig. 1 is a schematic flow chart of a model training method provided in an embodiment of the present specification, which may specifically include the following steps:

s100: for each commodity, commodity characteristics and historical data of the commodity are acquired.

The model training method provided by the specification is used for training the gain estimation model aiming at each commodity. The gain estimation model can predict the gain brought by pushing information corresponding to each commodity to the user. For example, taking push information as a donation traffic resource as an example, the gain estimation model may predict an increment of the transaction amount of a talk package due to the donation traffic to the user.

Therefore, in the present specification, when performing model training, the product characteristics and the history data of each product may be acquired first. The commodity characteristics at least comprise one of the category, the score, the price, the access amount and the order placing amount of the commodity, the historical data of the commodity at least comprise a first operated frequency and a second operated frequency of the commodity in the history, the first operated frequency refers to the frequency of performing the specified operation on the commodity by the user under the condition that the push information corresponding to the commodity is not pushed to the user, and the second operated frequency refers to the frequency of performing the specified operation on the commodity by the user under the condition that the push information corresponding to the commodity is pushed to the user.

In this specification, different definitions exist for commodities in different information push scenes, and taking push information as a presentation flow as an example, a commodity is a package service in different call durations. Taking the push information as a coupon for example, the goods may be various products to be sold. The specified operation executed by the user can be one or more of the behavior operations of accessing, placing orders, collecting, clicking, sharing and the like of the user.

S102: and determining the gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the total resource amount contained in the pushing information corresponding to the commodity historically pushed to the user.

In one or more embodiments of the present disclosure, after determining the history data of each product, the influence of the push information on the product may be determined based on the number of times that each product is operated when the user receives the push information and does not receive the push information.

Specifically, since the resource amounts contained in different pieces of push information are different, when the influence of the push information on a commodity is calculated, the total resource amount contained in the push information corresponding to the commodity pushed to the user can be determined according to the resource amount contained in the push information corresponding to the commodity historically pushed to the user. Taking the push information as a coupon as an example, the resource amount contained in the push information is the coupon amount of the coupon. Assuming that the push information corresponding to the product a is 10 coupons with a total of 100 to 15 coupons and 10 coupons with a total of 150 to 30 coupons, the total amount of resources included in the push information corresponding to the product a is 10 × 15+10 × 30, which is 450.

And then, determining the gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the total resource amount contained in the pushing information corresponding to the commodity historically pushed to the user.

Further, after the push information corresponding to the commodity is pushed to the user, part of the users may not actually use the resource corresponding to the push information. Therefore, when the gain of the unit resource amount with respect to the product is determined, the total amount of resources actually used by the user in the history can be determined from the total amount of resources included in the push information corresponding to the product which has been pushed to the user in the history. And then, determining the gain of the unit resource amount relative to the commodity according to the historical first operated times and second operated times of the commodity and the historical total resource amount actually used by the user.

For example, assuming that the coupons actually used by the user in the coupons of the product a are 4 coupons full of 100 to 15 coupons and 6 coupons full of 150 to 30 coupons, the total coupon amount actually used by the user is 15 × 4+10 × 6 — 120.

In an embodiment of the present specification, the first number of times of operation includes a first access amount and a first order amount of the user to the product when the push information corresponding to the product is not pushed to the user. The second operated times comprise a second access amount and a second order amount of the user to the commodity under the condition that the push information corresponding to the commodity is pushed to the user.

Still taking the example of pushing the coupon, the first number of times of operation includes a first amount of access to the product and a first amount of orders made by the user when the coupon for the product is not pushed to the user, and the second number of times of operation includes a second amount of access to the product and a second amount of orders made by the user when the coupon for the product is pushed to the user. And pushing the total amount of resources contained in the pushing information corresponding to the commodity, namely the total discount sum of the pushed coupons, to the user.

Therefore, when calculating the gain of the unit resource amount relative to the product, that is, the gain of the 1-element coupon relative to the product, the first order conversion rate of the product can be determined according to the first historical visit amount and the first order amount of the product. And determining a second ordering conversion rate of the commodity according to the second historical visit amount and the second ordering amount of the commodity. And then, determining the gain of the coupon sum 1 Yuan relative to the commodity according to the historical first order-placing conversion rate, second visit amount and the total coupon sum of the pushed coupons of the commodity.

The gain may be an increase in the order amount, an increase in profit, or the like, and any index representing the increase of the merchandise may be considered as the gain, which is not limited in the present specification.

Further, the gain of the unit resource amount relative to the commodity can be calculated by the following formula:

wherein, I tableThe gain of the unit resource amount relative to the commodity, namely the gain of the preferential amount of 1 yuan relative to the commodity is shown. n represents the total amount of resources contained in the push information corresponding to each pushed commodity, that is, the total coupon amount pushed. p is a radical of₁Denotes the first lower conversion, p₂Representing a second lower single conversion, and v represents a second visit volume.

It should be noted that, since the pushed coupon for the product is not actually used by all users, the gain of the actual unit discount amount with respect to the product is calculated more accurately. N in the above formula may also represent the total amount of the coupon for the goods actually used by the user.

S104: and taking the commodity characteristics of each commodity as each training sample, and taking the gain of the unit resource amount relative to each commodity as the label of each training sample.

S106: and aiming at each training sample, inputting the training sample into a gain estimation model to be trained, determining the gain to be optimized relative to the commodity by the unit resource quantity output by the gain estimation model, and adjusting model parameters in the gain estimation model to be trained by taking the minimized difference between the gain to be optimized and the label of the training sample as a target.

In one or more embodiments of the present specification, since the gain estimation model to be trained is used to estimate the influence of pushing information to a user on a commodity, after determining the gain of a unit resource amount in the pushing information on the commodity based on step S102, the determined gain may be used as a sample label to perform model training.

Specifically, commodity features of each commodity are used as training samples, and gains of unit resource amounts relative to the commodities are used as labels of the training samples. And then, aiming at each training sample, inputting the training sample into a gain estimation model to be trained, and determining the gain to be optimized, which is relative to the commodity, of the unit resource quantity output by the gain estimation model. And finally, adjusting model parameters in the gain estimation model to be trained by taking the minimized difference between each gain to be optimized and the label of each training sample as a target. And determining the information to be pushed corresponding to each commodity and pushing the information based on the gain brought by the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each current commodity based on the gain brought by the unit resource amount relative to each commodity and the trained gain estimation model. That is, the information pushing process shown in fig. 2, which is not described herein again. The gain estimation model may be a Decision Tree model, such as a Gradient Boosting Decision Tree (GBDT).

Based on the model training method shown in fig. 1, the gain of the unit resource amount for each commodity can be determined based on the historical data of each commodity and the total amount of resources included in the push information corresponding to the commodity historically pushed to the user. And determining each training sample and its label according to the commodity characteristics of each commodity and the gain brought by the allocation of unit resource amount. And then, inputting each training sample into a gain estimation model to be trained, and adjusting model parameters by taking the difference of the unit resource quantity output by the minimized model relative to the gain to be optimized brought by each commodity and the label of each training sample as a target. By training the gain estimation model, the gain of the unit resource amount of the pushed information relative to each commodity can be predicted, and information is pushed to the user based on the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity. The effectiveness of information pushing is improved, the resource amount corresponding to the pushed information is more in line with the user requirements, and the waste of the resources of the pushed information is reduced.

Based on the model training method shown in fig. 1, the present specification also provides an information push method, which can predict the gain of the unit resource amount of the push information relative to each commodity based on the gain estimation model trained by the method, and further push information to each user.

The technical solutions provided by the embodiments of the present application are described in detail below with reference to the accompanying drawings.

Fig. 2 is a schematic flow chart of an information pushing method provided in an embodiment of the present specification, which specifically includes the following steps:

s200: and acquiring the commodity characteristics of each commodity.

S202: and aiming at each commodity, inputting the commodity characteristics of the commodity into a pre-trained gain estimation model, and determining the gain of the unit resource quantity relative to the commodity.

When information push is performed in this specification, a gain estimation model obtained by training with the model training method may be used to predict gains of unit resource amounts relative to commodities, and then push information corresponding to commodities is determined based on a prediction result and a total resource amount included in information to be pushed, so as to perform information push.

The information pushing method can be executed by a server of an information pushing platform, the server can be an independent server or a system composed of a plurality of servers, such as a distributed server and the like, and can be a physical server or a cloud server.

Specifically, when pushing the push information corresponding to each commodity to the user, the commodity characteristics of each commodity may be acquired first. Wherein the commodity characteristics at least comprise one of the category, the score, the price, the visit amount and the order amount of the commodity. And then, aiming at each commodity, inputting the commodity characteristics of the commodity into a pre-trained gain estimation model, and determining the gain of the unit resource quantity relative to the commodity. The gain estimation model may be trained in advance by using the model training method, and the description is omitted here.

S204: determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and pushing the information to be pushed corresponding to each commodity to a user according to the type of the information to be pushed corresponding to each commodity.

In one or more embodiments of the present disclosure, since the total amount of resources of the information to be pushed to the user is limited, after the gain of the unit resource amount to each commodity is predicted by the gain prediction model, the amount of resources included in the pushed information corresponding to each commodity can be determined based on the gain of the unit resource amount to each commodity and the total amount of resources included in the information to be pushed.

Specifically, the type of the information to be pushed corresponding to each commodity is determined according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity. The resource amount corresponding to the push information of different types is different, and the total resource amount contained in the information to be pushed corresponding to each commodity is fixed. And then pushing the push information corresponding to each commodity to the user according to the determined type of the information to be pushed corresponding to each commodity. The larger the gain of the unit resource amount relative to the commodity is, the more the resource amount of the type of the information to be pushed corresponding to the commodity is.

Because the total amount of resources contained in the information to be pushed for providing services for the user is limited, when the type of the information to be pushed corresponding to each commodity is determined, the total order amount or the total transaction amount of each commodity can be maximized, and the amount of the resources in the information to be pushed corresponding to each commodity is determined according to the total amount of the resources contained in the information to be pushed corresponding to each commodity and the gain of the unit resource amount relative to each commodity.

Further, the server may divide the plurality of commodity groups in accordance with the ranking of the unit resource amount with respect to the gain due to each commodity. And then, determining the type of the information to be pushed corresponding to each commodity group according to the quantity of the commodities in each commodity group and the total quantity of resources contained in the information to be pushed corresponding to each commodity. And finally, determining the type of the information to be pushed corresponding to each commodity according to the type of the information to be pushed corresponding to each commodity group. The types of the information to be pushed corresponding to the commodities in the same commodity group are the same.

Of course, the server may also group the commodities according to a preset group gain threshold, which may be set as needed.

Furthermore, in order to improve the effectiveness of information push, information push can be performed in combination with user personalized information, such as the sensitivity of the user on whether to push information. Therefore, for each user, the server can also determine the ordering probability of the user under the condition that the user does not receive the push information corresponding to the commodity and the ordering probability of the user under the condition that the user receives the push information corresponding to the commodity according to the portrait information and the historical behavior information of the user. And then, pushing the push information corresponding to the commodity to the user according to the ordering probability of the user under the condition that the user does not receive the push information corresponding to the commodity, the ordering probability of the user under the condition that the user receives the push information corresponding to the commodity and the type of the push information corresponding to the commodity.

When the ordering probability of the user is high under the condition that the user does not receive the push information corresponding to the commodity, the user can be determined to be insensitive to the resource acquisition service based on the push information, and the push information corresponding to the commodity can not be pushed to the user. When the ordering probability of the user is high under the condition that the user receives the push information corresponding to the commodity, the user can be determined to be sensitive to the resource acquisition service based on the push information, and the push information corresponding to the commodity can be pushed to the user according to the type of the push information corresponding to the commodity.

Based on the information push method shown in fig. 2, the gain of the unit resource amount for each commodity can be determined by a gain estimation model trained in advance based on the commodity characteristics of each commodity. And then, determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount contained in the information to be pushed corresponding to each commodity, and pushing the information to the user. And estimating the gain of the unit resource quantity relative to each commodity through a gain estimation model, and determining the push information corresponding to each commodity pushed to the user according to the gain of the unit resource quantity relative to each commodity. The effectiveness of information pushing is improved, the resource amount corresponding to the pushed information is more in line with the user requirements, and the waste of the resources of the pushed information is reduced.

In an embodiment of the present specification, the information push method may be applied to an e-commerce platform to push coupon information. The goods may be goods for sale in the e-commerce platform.

Specifically, the e-commerce platform can determine each commodity to be set with a discount and acquire the commodity characteristics of each commodity. And then, aiming at each commodity, inputting the commodity characteristics of the commodity into a pre-trained gain estimation model, and determining the gain of the unit preferential amount relative to the commodity. The gain represents the gain brought by issuing the coupon of the unit discount sum, and can be the increase of the commodity order amount or the increase of the commodity bargain amount. When the gain is characterized as an increase in the order amount for the good, then the gain in the unit offer amount relative to the good may be characterized as an order increment due to the 1-coupon for the good being issued to the user.

After determining the gain of the unit preferential amount relative to each commodity, the E-commerce platform can divide a plurality of commodity groups according to the sequence of the unit preferential amount relative to each commodity, and determine the type of the coupon corresponding to each commodity group according to the quantity of the commodities in each commodity group and the total preferential amount of the coupon to be pushed. Wherein the offer amounts in different types of coupons differ.

For example, the divided commodities are grouped into 3 groups, and the groups are respectively a high gain group, a medium gain group and a low gain group according to the gain high-low order. Wherein, the high gain group comprises 5 commodities, the total commodity quantity is 100, the medium gain group comprises 8 commodities, the total commodity quantity is 400, the low gain group comprises 10 commodities, the total commodity quantity is 500, and if the total coupon sum to be pushed is 2 ten thousand yuan at most, it can be determined that 200-25 coupons can be issued for each commodity in the high gain group, 200-20 coupons can be issued for each commodity in the medium gain group, and 200-15 coupons can be issued for each commodity in the low gain group.

Furthermore, the e-commerce platform can also determine the ordering probability of each user to be pushed under different preferential amounts according to the portrait information and the historical behavior information of the user. And determining the coupon amount for purchasing the commodity, which is issued to the user, according to the ordering probability of the user under different coupon amounts and the type of the push information corresponding to the commodity.

Based on the model training method shown in fig. 1, an embodiment of the present specification further provides a schematic structural diagram of a model training apparatus, as shown in fig. 3.

Fig. 3 is a schematic structural diagram of a model training apparatus provided in an embodiment of the present disclosure, including:

the acquisition module 300 is configured to acquire, for each commodity, a commodity feature of the commodity and history data, where the history data includes at least a first number of times of operation and a second number of times of operation of the commodity in history; the first operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is not pushed to the user; the second operated times are times of executing specified operations on the commodity by the user under the condition that the push information corresponding to the commodity is pushed to the user;

a determining module 302, configured to determine, according to the historical first operated frequency and second operated frequency of the commodity, and a total amount of resources included in push information corresponding to the commodity historically pushed to a user, a gain brought by a unit resource amount relative to the commodity;

a sample preparation module 304, configured to use the commodity features of each commodity as each training sample, and use the gain of the unit resource amount relative to each commodity as the label of each training sample;

the training module 306 is configured to input the training sample into a gain estimation model to be trained for each training sample, determine a gain to be optimized, which is output by the gain estimation model and is caused by the unit resource amount relative to the commodity, and adjust a model parameter in the gain estimation model to be trained with a goal of minimizing a difference between the gain to be optimized and a label of the training sample, wherein the gain estimation model is used for estimating gains caused by the unit resource amount relative to the commodities, and determines information to be pushed corresponding to the commodities and pushes the information based on the gains caused by the unit resource amount relative to the commodities and a total resource amount included in the information to be pushed corresponding to the commodities.

Optionally, the determining module 302 is specifically configured to determine, from the total amount of resources included in the pushing information corresponding to the commodity historically pushed to the user, the total amount of resources actually used by the user historically, and determine, according to the first number of times of operation and the second number of times of operation of the commodity historically and the total amount of resources actually used by the user historically, a gain brought by the unit resource amount relative to the commodity.

Optionally, the first number of times of operation includes a first access amount and a first order amount of the user to the product under the condition that the push information corresponding to the product is not pushed to the user, the second operated number of times includes a second access amount and a second order amount of the user to the product when pushing the push information corresponding to the product to the user, where the determining module 302 is specifically configured to, determining a first order conversion rate of the commodity according to the first historical visit amount and the first order amount of the commodity, determining a second ordering conversion rate of the commodity according to a second historical visit amount and a second ordering amount of the commodity, and determining the gain of the unit resource amount relative to the commodity according to the first ordering conversion rate, the second visit amount and the total amount of resources contained in push information corresponding to the commodity historically pushed to the user.

Based on the information pushing method shown in fig. 2, an embodiment of the present specification further provides a schematic structural diagram of an information pushing apparatus, as shown in fig. 4.

Fig. 4 is a schematic structural diagram of an information pushing apparatus provided in an embodiment of the present specification, including:

an obtaining module 400 configured to obtain the commodity characteristics of each commodity;

a determining module 402, configured to input, for each commodity, the commodity characteristics of the commodity into a pre-trained gain estimation model, and determine a gain brought by a unit resource amount with respect to the commodity; the gain estimation model is obtained by training by adopting the model training method;

the allocation module 404 is configured to determine the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total amount of resources included in the information to be pushed corresponding to each commodity, and push the information to be pushed corresponding to each commodity to the user according to the type of the information to be pushed corresponding to each commodity, where the resource amounts corresponding to different types of push information are different.

Optionally, the allocating module 404 is specifically configured to divide a plurality of commodity groups according to the sorting of the unit resource amount relative to the gain brought by each commodity, determine the type of information to be pushed corresponding to each commodity group according to the quantity of commodities in each commodity group and the total resource amount included in the information to be pushed corresponding to each commodity, and determine the type of information to be pushed corresponding to each commodity according to the type of information to be pushed corresponding to each commodity group; the types of the information to be pushed corresponding to the commodities in the same commodity group are the same.

Optionally, the allocating module 404 is further configured to, for each user, determine, according to the portrait information and the historical behavior information of the user, an ordering probability of the user in a case where the user does not receive the push information corresponding to the product, and determine, in a case where the user receives the push information corresponding to the product, an ordering probability of the user in a case where the user does not receive the push information corresponding to the product, an ordering probability of the user in a case where the user receives the push information corresponding to the product, and a type of the push information corresponding to the product, to push the push information corresponding to the product to the user.

Embodiments of the present specification further provide a computer-readable storage medium, where the storage medium stores a computer program, and the computer program may be used to execute the model training method provided in fig. 1 or the information pushing method provided in fig. 2.

According to the model training method shown in fig. 1 or the information pushing method shown in fig. 2, an embodiment of the present specification further provides a schematic structural diagram of the electronic device shown in fig. 5. As shown in fig. 5, 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 method shown in fig. 1 or the information pushing method shown in fig. 2.

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 create a dedicated integrated circuit chip. Furthermore, nowadays, instead of manually generating 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 Language Description Language), traffic, pl (core unified Programming Language), HDCal, JHDL (Java Hardware Description Language), langue, Lola, HDL, laspam, hardbyscript Description Language (vhigh Description Language), and so on, which are currently used in the most popular languages. 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:

2. The method according to claim 1, wherein determining a gain of a unit resource amount relative to the commodity according to a historical first operated frequency and a historical second operated frequency of the commodity and a total resource amount included in push information corresponding to the commodity historically pushed to a user, specifically comprises:

3. The method of claim 1, wherein the first number of times of operation includes a first amount of access to the product and a first amount of orders made by the user when the push information corresponding to the product is not pushed to the user, and the second number of times of operation includes a second amount of access to the product and a second amount of orders made by the user when the push information corresponding to the product is pushed to the user;

4. An information pushing method, comprising:

acquiring commodity characteristics of each commodity;

inputting the commodity characteristics of each commodity into a pre-trained gain estimation model, and determining the gain of unit resource quantity relative to the commodity; wherein the gain estimation model is obtained by training by adopting the method of any one of claims 1 to 3;

5. The method according to claim 4, wherein determining the type of the information to be pushed corresponding to each commodity according to the gain of the unit resource amount relative to each commodity and the total resource amount included in the information to be pushed corresponding to each commodity specifically comprises:

6. The method of claim 4, wherein the method further comprises:

7. A model training apparatus, comprising:

8. An information pushing apparatus, comprising:

the determining module is configured to input the commodity characteristics of each commodity into a pre-trained gain estimation model and determine the gain of the unit resource quantity relative to the commodity; wherein the gain estimation model is obtained by training by adopting the method of any one of claims 1 to 3;

9. 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 3 or 4 to 6.

10. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the method of any of claims 1 to 3 or 4 to 6.