CN114817716A

CN114817716A - Method, device, equipment and medium for predicting user conversion behaviors and training model

Info

Publication number: CN114817716A
Application number: CN202210416056.XA
Authority: CN
Inventors: 文豪
Original assignee: Baidu Online Network Technology Beijing Co Ltd
Current assignee: Baidu Online Network Technology Beijing Co Ltd
Priority date: 2022-04-20
Filing date: 2022-04-20
Publication date: 2022-07-29

Abstract

The present disclosure relates to the field of data processing technologies, and in particular, to the field of deep learning. The specific implementation scheme is as follows: acquiring user characteristics and resource characteristics; predicting the occurrence probability of a first conversion behavior based on the user characteristics and the resource characteristics, wherein the occurrence probability of the first conversion behavior is the probability of a user with the user characteristics performing the first conversion behavior on the resource corresponding to the resource characteristics; the first conversion behavior is associated with the second conversion behavior and is a previous operation behavior of the second conversion behavior, and the second conversion behavior is a preset target user conversion behavior needing to be predicted to occur. The user conversion behavior can be predicted more accurately through the method and the device.

Description

Method, device, equipment and medium for predicting user conversion behaviors and training model

Technical Field

The disclosure relates to the technical field of data processing, in particular to the field of resource recommendation, and specifically relates to a user transformation behavior prediction scene in the field of resource recommendation.

Background

With the development of science and technology, the requirement for the accuracy of resource recommendation is higher and higher in a resource recommendation scene.

In the related art, a user matching a resource is determined by predicting a probability that the user performs a conversion action on the resource (for example, the probability that the resource is clicked and used by the user (i.e., conversion probability)), and the resource is recommended. In order to be able to more accurately match user features to resources, a user translation behavior prediction model is applied.

When the user transformation behavior prediction model is trained, model prediction needs to be performed according to historical data (transformation samples) of transformation behaviors executed by the user, transformation behaviors which are possibly generated by the user are predicted based on the predicted model, and resource recommendation is performed.

Disclosure of Invention

The present disclosure provides a method, apparatus, device, storage medium and program product for predicting user conversion behavior, training a user conversion behavior prediction model.

According to an aspect of the present disclosure, there is provided a method of predicting user conversion behavior, including:

acquiring user characteristics and resource characteristics;

predicting a first conversion behavior occurrence probability based on the user characteristics and the resource characteristics, wherein the first conversion behavior occurrence probability is the probability of a user with the user characteristics performing a first conversion behavior on resources corresponding to the resource characteristics;

the first conversion behavior is associated with a second conversion behavior and is a previous operation behavior of the second conversion behavior, and the second conversion behavior is a target user conversion behavior for resource recommendation.

In an exemplary embodiment, predicting a first transition behavior occurrence probability based on the user characteristics and the resource characteristics includes:

inputting the user characteristics and the resource characteristics into a first user transformation behavior prediction model;

determining a first conversion behavior occurrence probability based on an output of the first user conversion behavior prediction model;

the first user conversion behavior prediction model is obtained by pre-training based on first sample data, wherein the first sample data comprises user characteristics, resource characteristics and a first label, and the first label is used for indicating that a user performs a first conversion behavior on resources corresponding to the resource characteristics;

the input of the first user conversion behavior prediction model is user characteristics and resource characteristics, and the output is the occurrence probability of the first conversion behavior.

inputting the user characteristics and the resource characteristics into a third user conversion behavior prediction model;

determining a first conversion behavior occurrence probability based on an output of the third user conversion behavior prediction model;

the third user conversion behavior prediction model is obtained by pre-training based on third sample data;

the third sample data comprises user characteristics, resource characteristics, a first label, a second label and resource use duration information, wherein the first label is used for indicating that a user performs a first conversion behavior on the resource corresponding to the resource characteristics, and the second label is used for indicating that the user performs a second conversion behavior on the resource corresponding to the resource characteristics;

and the input of the third user conversion behavior prediction model is user characteristics and resource characteristics, and the output is the occurrence probability of the first conversion behavior, the occurrence probability of the second conversion behavior and the resource use duration.

In an exemplary embodiment, the first conversion behavior occurrence probability and the second conversion behavior occurrence probability are weighted to obtain a target probability of the target user conversion behavior.

According to another aspect of the present disclosure, there is provided a method of training a user conversion behavior prediction model, including:

acquiring first sample data, wherein the first sample data comprises user characteristics, resource characteristics and a first label, the first label is used for indicating that a user carries out a first conversion behavior on a resource corresponding to the resource characteristics, and the first conversion behavior is associated with a second conversion behavior and is a previous operation behavior of the second conversion behavior;

inputting the user characteristics and the resource characteristics in the first sample data into a first prediction model as input characteristics to obtain a first conversion behavior occurrence probability prediction value;

training the first prediction model based on the first label and the first conversion behavior occurrence probability prediction value until the first prediction model converges to obtain a first user conversion behavior prediction model;

the output of the first user conversion behavior prediction model is a first conversion behavior occurrence probability.

determining second sample data, wherein the second sample data comprises a user characteristic, a resource characteristic, a first label and a second label, the first label is used for indicating that a user carries out a first conversion behavior on a resource corresponding to the resource characteristic, and the second label is used for indicating that the user carries out a second conversion behavior on the resource corresponding to the resource characteristic;

inputting the user characteristics and the resource characteristics in the second sample data as input characteristics into a second prediction model to obtain a first conversion behavior occurrence probability prediction value and a second conversion behavior occurrence probability prediction value;

training the second prediction model based on the first conversion behavior occurrence probability prediction value and the first label, and based on the second conversion behavior occurrence probability prediction value and the second label to obtain a second user conversion behavior prediction model;

and the output of the second user conversion behavior prediction model is a first conversion behavior occurrence probability and a second conversion behavior occurrence probability.

determining third sample data, wherein the third sample data comprises user characteristics, resource characteristics, a first label and a second label as well as resource use duration information, the first label is used for indicating that a user performs a first conversion behavior on a resource corresponding to the resource characteristics, and the second label is used for indicating that the user performs a second conversion behavior on the resource corresponding to the resource characteristics;

inputting the user characteristics and the resource characteristics in the third sample data as input characteristics into a third prediction model to obtain a first conversion behavior occurrence probability prediction value, a second conversion behavior occurrence probability prediction value and a resource duration prediction value;

training the third prediction model based on the first conversion behavior occurrence probability prediction value and the first label, the second conversion behavior occurrence probability prediction value and the second label, and the resource duration prediction value and the resource use duration information to obtain a third user conversion behavior prediction model;

and the output of the third user conversion behavior prediction model is the occurrence probability of the first conversion behavior, the occurrence probability of the second conversion behavior and the resource use duration.

According to another aspect of the present disclosure, there is provided an apparatus for predicting user conversion behavior, including:

the device comprises an acquisition unit, a processing unit and a processing unit, wherein the acquisition unit is used for acquiring user characteristics and resource characteristics;

the prediction unit is used for predicting a first conversion behavior occurrence probability according to the user characteristics and the resource characteristics, wherein the first conversion behavior occurrence probability is a probability that a user with the user characteristics performs a first conversion behavior on a resource corresponding to the resource characteristics, the first conversion behavior is associated with a second conversion behavior and is a previous operation behavior of the second conversion behavior, and the second conversion behavior is a preset target user conversion behavior needing to be predicted.

In an exemplary embodiment, the prediction unit is configured to predict the occurrence probability of the first transition behavior based on the user characteristic and the resource characteristic in the following manner:

In an exemplary embodiment, the prediction unit is configured to predict the occurrence probability of the first conversion behavior based on the user characteristic and the resource characteristic in the following manner:

inputting the user characteristics and the resource characteristics into a second user transformation behavior prediction model;

determining a first conversion behavior occurrence probability based on an output of the second user conversion behavior prediction model;

the second user conversion behavior prediction model is obtained by pre-training based on second sample data;

the second sample data comprises user characteristics, resource characteristics, a first label and a second label, the first label is used for indicating that a user carries out a first conversion behavior on the resource corresponding to the resource characteristics, and the second label is used for indicating that the user carries out a second conversion behavior on the resource corresponding to the resource characteristics;

and the input of the second user conversion behavior prediction model is user characteristics and resource characteristics, and the output is the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior.

In an exemplary embodiment, the prediction unit is further configured to:

and weighting the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior to obtain a target probability of the target user conversion behavior.

According to another aspect of the present disclosure, there is provided a device for training a user conversion behavior prediction model, including:

the device comprises an obtaining unit, a processing unit and a processing unit, wherein the obtaining unit is used for obtaining first sample data, the first sample data comprises user characteristics, resource characteristics and a first label, the first label is used for indicating that a user carries out a first conversion behavior on a resource corresponding to the resource characteristics, the first conversion behavior is associated with a second conversion behavior, and the first conversion behavior is a previous operation behavior of the second conversion behavior;

the prediction unit is used for inputting the user characteristics and the resource characteristics in the first sample data into a first prediction model as input characteristics to obtain a first conversion behavior occurrence probability prediction value;

and the training unit is used for training the first prediction model according to the first label and the first conversion behavior occurrence probability prediction value until the first prediction model converges to obtain a first user conversion behavior prediction model, and the output of the first user conversion behavior prediction model is the first conversion behavior occurrence probability.

According to another aspect of the present disclosure, there is provided an apparatus for training a user conversion behavior prediction model, including:

the determining unit is configured to determine second sample data, where the second sample data includes a user characteristic, a resource characteristic, a first tag and a second tag, the first tag is used to indicate that a user performs a first conversion behavior on a resource corresponding to the resource characteristic, and the second tag is used to indicate that the user performs a second conversion behavior on the resource corresponding to the resource characteristic;

the prediction unit is used for inputting the user characteristics and the resource characteristics in the second sample data into a second prediction model as input characteristics to obtain a first conversion behavior occurrence probability prediction value and a second conversion behavior occurrence probability prediction value;

and the training unit is used for training the second prediction model according to the first conversion behavior occurrence probability prediction value and the first label and based on the second conversion behavior occurrence probability prediction value and the second label to obtain a second user conversion behavior prediction model, and the output of the second user conversion behavior prediction model is the first conversion behavior occurrence probability and the second conversion behavior occurrence probability.

According to another aspect of the present disclosure, there is provided a training user conversion behavior prediction apparatus, including:

the determining unit is configured to determine third sample data, where the third sample data includes a user characteristic, a resource characteristic, a first tag, a second tag, and resource usage duration information, the first tag is used to indicate that a user performs a first conversion behavior on a resource corresponding to the resource characteristic, and the second tag is used to indicate that the user performs a second conversion behavior on the resource corresponding to the resource characteristic;

the prediction unit is used for inputting the user characteristics and the resource characteristics in the third sample data into a third prediction model to obtain a first conversion behavior occurrence probability prediction value and a second conversion behavior occurrence probability prediction value;

and the training unit is used for training the third prediction model according to the first conversion behavior occurrence probability prediction value, the first label, the second conversion behavior occurrence probability prediction value, the second label, the resource use duration prediction value and the resource use duration information to obtain a third user conversion behavior prediction model, and the output of the third user conversion behavior prediction model is the first conversion behavior occurrence probability, the second conversion behavior occurrence probability and the resource use duration.

According to another aspect of the present disclosure, there is provided an electronic device including:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method described above.

According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the above-described method.

According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements the method described above.

The method and the device for predicting the user conversion behavior more accurately are provided through the disclosure.

It should be understood that the statements in this section do not necessarily identify key or critical features of the embodiments of the present disclosure, nor do they limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description.

Drawings

The drawings are included to provide a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 is a diagram illustrating a relationship between a user click rate and a conversion rate;

FIG. 2 is a flowchart illustrating a method of predicting user conversion behavior in accordance with an exemplary embodiment of the present disclosure;

FIG. 3 is a schematic flow chart diagram of a method for predicting the occurrence probability of a first transition behavior of a user according to the present disclosure;

FIG. 4 is a schematic flow chart diagram illustrating a method for predicting a probability of occurrence of a first transition behavior of a user according to the present disclosure;

FIG. 5 is a schematic flow chart diagram illustrating a method for predicting a probability of occurrence of a first transition behavior of a user according to the present disclosure;

FIG. 6 is a schematic flow chart illustrating a method for predicting the occurrence probability of a user transformation behavior according to the method for predicting a user transformation behavior provided by the present disclosure;

FIG. 7 is a schematic flow diagram of training a first user conversion behavior prediction model training provided in accordance with the present disclosure;

FIG. 8 is a schematic flow diagram of training a second user conversion behavior prediction model training provided in accordance with the present disclosure;

FIG. 9 is a schematic diagram of a training architecture for a second user translation behavior prediction model provided in accordance with the present disclosure;

FIG. 10 is a diagram illustrating a correspondence between a resource usage duration and a conversion behavior occurrence probability of a user;

FIG. 11 is a schematic flow diagram of training a third user translation behavior prediction model provided in accordance with the present disclosure;

FIG. 12 illustrates a network architecture diagram of a third user translation behavior prediction model, shown in an exemplary embodiment of the present disclosure;

FIG. 13 is a block diagram of an apparatus for predicting user translation behavior in accordance with an exemplary embodiment of the present disclosure;

FIG. 14 is a block diagram of an apparatus for training a predictive model of user translation behavior provided in accordance with the present disclosure;

FIG. 15 is a block diagram of an apparatus for training a predictive model of user translation behavior provided in accordance with the present disclosure;

FIG. 16 is a block diagram of an apparatus for training a predictive model of user translation behavior provided in accordance with the present disclosure;

fig. 17 illustrates a schematic block diagram of an example electronic device 1700 that can be used to implement embodiments of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, in which various details of the embodiments of the disclosure are included to assist understanding, and which are to be considered as merely exemplary. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The method for predicting the user transformation behavior can be applied to resource recommendation scenes. For example, the method can be applied to resource recommendation in a multi-element presentation scene. The resource recommendation method can be applied to resource recommendation in a multivariate presentation scene of an application program (APP) and a service platform (Feed).

Wherein, accurate recommendation of resources is crucial to users and platforms. For example, in a multi-occurrence scenario, multi-occurrence resources are required to be distributed over the Feed of the application. The multiple resource change refers to some resources with user payment behaviors, such as payment resources and delivery resources. The resources are distributed on Feed, so that more content showing channels can be provided for an author, in addition to advertising income, content paying income can be provided for the author, and the initiative of author document sending is promoted; for the user, the quality of the multi-element cash resources is higher than that of common resources, and the user can meet the requirements of user knowledge acquisition and commodity purchase through the multi-element cash resources; for the platform, the distribution of multiple emerging resources on the Feed platform is an important ring of the electric business strategy of the company, and the electric business strategy can be assisted to develop.

In the related art, resources which are interested by a user are recommended for the user according to the occurrence probability of the user conversion behavior aiming at resource recommendation. The user conversion behavior refers to a preset target user conversion behavior generated when a user clicks a target object. The user conversion behavior occurrence probability refers to the probability that a preset target user conversion behavior occurs after a user clicks a target object. The target object can be understood as a resource displayed when resource recommendation is performed. For example, the target object may be a picture of a commodity, a picture of APP, etc., and is used for advertising purposes, i.e., a commodity advertisement. The target object is not limited to this. For example, in the multi-occurrence scenario, the user conversion behavior may be a behavior in which the user clicks on a multi-occurrence resource such as a merchandise card or a shopping cart, and makes a purchase.

In the resource recommendation field, the occurrence probability of the user conversion behavior can be predicted through a pre-trained user conversion behavior prediction model. Wherein the prediction of the user conversion behavior relies on a large number of training samples. The training sample is mainly determined based on user conversion behavior data generated by user historical operation. However, there is a problem of sparse training samples. For example, in the field of multi-change-occurrence recommendation, the largest difference between multi-change-occurrence resources and common resources is that the multi-change-occurrence resources have attributes of user paid conversion, and the occurrence probability of the display amount of the multi-change-occurrence resources, the click rate of the multi-change-occurrence resources and the conversion behavior of the user is greatly different, for example, the display amount of the multi-change-occurrence resources per day is in the hundred million level, the click rate is in the ten million level, and the conversion amount is only in the ten million level. Therefore, there is a problem that the user conversion behavior is sparse. How to use the sparse user transformation behaviors enables the user transformation behavior prediction model to obtain an accurate prediction result so as to accurately recommend the resources which are most interested in the user and have the highest probability of the user transformation behaviors for the user, which is a key technology.

In the related art, in order to solve the problem of sparse user conversion behaviors, one mode is to jointly model according to a user click behavior and a conversion behavior, jointly model the user conversion behavior and the user click behavior in a display sample space, and train the conversion behavior by using the user click behavior to assist in the conversion behavior, so that the problem of sparse user conversion behaviors is solved. Another way is to build the model by accumulating samples, i.e. by performing the accumulation of samples with user conversion behavior over a long time.

The method is mainly based on the assumption that the user click behavior and the user purchase behavior are related, namely, the user clicking the resource is more likely to generate the purchase behavior. However, in the practical application process, the relationship between the user click rate and the conversion rate on the multivariate demonstration resource is calculated, so that the correlation relationship between the user click rate and the conversion rate is not obvious. Fig. 1 shows a schematic diagram of a correspondence relationship between a user click rate and a conversion rate. Referring to fig. 1, as the conversion rate of the user increases, the click rate of the user does not change significantly, which may indicate that the relationship between the click and the conversion behavior of the user is not significant. In addition, in practical application, according to a mode of jointly modeling the user click behavior and the conversion behavior, the estimation accuracy of the user conversion behavior prediction model is not obviously improved.

For the way of accumulating samples, the conversion behavior of the user can be accumulated to the level available for the model for training. But at the same time will also have an influence on the effect of the model. Because the user conversion behavior prediction model is basically trained and updated in the hour or minute level, the behavior of the user in a short period can be quickly added into the model training, and the estimation feedback of the model is obtained. The method of accumulating samples prolongs the updating time of the user conversion behavior model, so that the short-term behavior samples of the user cannot be added into model training, the model cannot finely describe the short-term behavior of the user, and the estimation accuracy cannot be guaranteed.

In view of this, the present disclosure provides a method for predicting user transformation behavior. In the method, the operation behavior of the previous step of the target user conversion behavior for resource recommendation is predicted.

For convenience in description in the embodiments of the present disclosure, a previous operation behavior of a target user conversion behavior for resource recommendation is referred to as a first conversion behavior, and is sometimes referred to as a shallow conversion behavior. The target user conversion behavior for resource recommendation is referred to as a second conversion behavior, sometimes also referred to as a deep conversion behavior. Wherein the first conversion behavior is associated with the second conversion behavior and is a previous operation behavior in which the second conversion behavior occurs. For example, in the scenario of resource recommendation of multiple renditions, the first transition behavior may be a user clicking a shopping cart, clicking a purchase button, clicking a merchandise card, and so on. The second conversion activity may also be a purchase activity by the user.

According to the method for predicting the user transformation behaviors, the occurrence probability of the first transformation behaviors is predicted, and the probability of the second transformation behaviors is higher under the condition that the first transformation behaviors occur because the first transformation behaviors are the previous operation behaviors of the second transformation behaviors, so that the second transformation behaviors can be simulated through the first transformation behaviors in the embodiment of the disclosure, the second transformation behaviors are enriched, and the problem of sparseness of the second transformation behaviors is solved.

The present disclosure will now be described with reference to a method for predicting user conversion behavior in exemplary embodiments of the present disclosure.

As an exemplary embodiment, FIG. 2 is a flow chart illustrating a method for predicting user conversion behavior according to an exemplary embodiment of the present disclosure. Referring to fig. 2, the method for predicting user transformation behavior provided by the present disclosure includes the following steps.

In step S201, a user characteristic and a resource characteristic are acquired.

In step S202, a first conversion behavior occurrence probability is predicted based on the user characteristics and the resource characteristics.

In the present disclosure, the user characteristics include basic attributes of the user, such as age, gender, and the like. The resource characteristics comprise basic attributes of the resources, such as exposure time periods of the resources, materials of the resources and the like; the resource characteristics may also include scene characteristic information, such as basic attributes of an exposure scene of the exposed resource, such as what web page the exposed resource is on, what application program triggers exposure, and the like. And the user characteristics and the resource characteristics have corresponding matching relations.

In the disclosure, the occurrence probability of the first conversion behavior is the probability of the user with the user characteristic performing the first conversion behavior on the resource corresponding to the resource characteristic, the first conversion behavior is associated with the second conversion behavior and is a previous operation behavior of the second conversion behavior, and the second conversion behavior is a preset target user conversion behavior which needs to be predicted to occur.

As an exemplary embodiment, in the scenario of resource recommendation, the user characteristic may be user data such as a behavior of a user clicking a shopping cart, a behavior of a user clicking a resource card, a user gender, a user name of the user, a browsing duration, and a user history conversion, and the resource characteristic may be resource type information, such as a type of live content.

In the embodiment of the disclosure, after the user characteristics and the resource characteristics are obtained, the probability of the conversion behavior of the user on the currently browsed page, that is, the probability of the first conversion behavior is predicted.

According to the method for predicting the user transformation behavior, the probability of occurrence of the first transformation behavior is predicted according to the user characteristics and the resource characteristics acquired by the method. The first transformation behavior is a previous operation behavior of the second transformation behavior, so that the probability of the second transformation behavior is higher under the condition that the first transformation behavior occurs, and therefore in the embodiment of the disclosure, the second transformation behavior can be simulated through the first transformation behavior to enrich the second transformation behavior and improve the problem of sparseness of the second transformation behavior. In addition, in the embodiment of the disclosure, resource recommendation can be performed with the probability of occurrence of the first conversion behavior, so that the conversion behavior types of resource recommendation are enriched, the accuracy of predicting the user conversion behavior is higher, and the association between the user characteristics and the resource characteristics is tighter.

In an exemplary embodiment of the present disclosure, when predicting the occurrence probability of the first conversion behavior based on the user characteristics and the resource characteristics, the prediction may be performed based on a first user conversion behavior prediction model trained in advance. In the present disclosure, the input of the first user conversion behavior prediction model is a user characteristic and a resource characteristic, and the output is a first conversion behavior occurrence probability.

As an exemplary embodiment, fig. 3 is a flowchart illustrating a method for predicting an occurrence probability of a first transition behavior of a user according to the present disclosure. Referring to fig. 3, the method for predicting the occurrence probability of the first transition behavior of the user provided by the present disclosure includes the following steps.

In step S301, the user characteristics and the resource characteristics are input as input characteristics to the first user conversion behavior prediction model.

In step S302, a first conversion behavior occurrence probability is determined based on an output of the first user conversion behavior prediction model.

And the first user conversion behavior prediction model is obtained by pre-training based on the first sample data. The first sample data comprises user characteristics, resource characteristics and a first label, and the first label is used for indicating that a user carries out first conversion behaviors on resources corresponding to the resource characteristics.

In the method for predicting the user conversion behavior provided by the embodiment of the disclosure, the probability of the first conversion behavior can be obtained by obtaining a first user conversion behavior prediction model through pre-training based on the first sample data to predict the occurrence probability of the first conversion behavior. The first user transformation behavior prediction model is trained based on user characteristics, resource characteristics and a first label for identifying the first transformation behavior, and the number of the second transformation behaviors can be enriched due to the first transformation behavior, so that the prediction accuracy is higher compared with the user transformation behavior prediction model obtained by training based on the second transformation behavior, and the probability prediction of the first transformation behavior is more accurate.

In an exemplary embodiment of the present disclosure, when predicting the occurrence probability of the first conversion behavior based on the user characteristics and the resource characteristics, the prediction may be performed based on a second user conversion behavior prediction model trained in advance. In the disclosure, the input of the second user conversion behavior prediction model is the user characteristic and the resource characteristic, and the output is the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior.

As an exemplary embodiment, fig. 4 is a flowchart illustrating a method for predicting an occurrence probability of a first transition behavior of a user according to the present disclosure. Referring to fig. 4, the method for predicting the occurrence probability of the first transition behavior of the user provided by the present disclosure includes the following steps.

In step S401, the user characteristics and the resource characteristics are input as input characteristics to the second user conversion behavior prediction model.

In step S402, a first conversion behavior occurrence probability is determined based on an output of the second user conversion behavior prediction model.

In the disclosure, the second user transformation behavior prediction model is obtained by pre-training based on second sample data, where the second sample data includes a user characteristic, a resource characteristic, a first label and a second label. The first label is used for indicating that a user carries out a first conversion action on the resource corresponding to the resource characteristic, and the second label is used for indicating that the user carries out a second conversion action on the resource corresponding to the resource characteristic.

In the embodiment of the disclosure, the training sample of the second user conversion behavior prediction model contains the first sample data, and the first sample data has an effect of assisting in training the second user conversion behavior prediction model in determining the occurrence probability of the second conversion behavior, so that the second user conversion behavior prediction model is more accurate, and the problem that the second sample data is unavailable or sparse is solved.

The auxiliary training can be understood as enriching the second user transformation behavior prediction model by using the first sample data, and the second sample data is sparse, so that the data finally transformed by the user is rare in quantity compared with the user characteristic data, the first sample data is also included in the training of the second user transformation behavior prediction model, the second user transformation behavior prediction model can be matched with the resource characteristics more, and samples required by the model are not required to be accumulated for a long time.

In an embodiment of the disclosure, the output of the second user conversion behavior prediction model includes a first conversion behavior occurrence probability and a second conversion behavior occurrence probability. Therefore, in the embodiment of the present disclosure, resource recommendation may be performed based on the first conversion behavior occurrence probability and the second conversion behavior occurrence probability.

In an exemplary embodiment of the present disclosure, the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior may be weighted to obtain a target probability of the occurrence of the conversion behavior of the target user, and resource recommendation is performed based on the target probability.

In the embodiment of the disclosure, resource recommendation is performed based on the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior, the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior are fused, and resource recommendation is performed by using the second conversion behavior with sparse dependence number, so that the accuracy of resource recommendation can be improved.

In an exemplary embodiment of the present disclosure, when predicting the occurrence probability of the first conversion behavior based on the user characteristics and the resource characteristics, the prediction may be performed based on a third user conversion behavior prediction model trained in advance. In the present disclosure, the input of the third user conversion behavior prediction model is the user characteristic and the resource characteristic, and the output is the occurrence probability of the first conversion behavior, the occurrence probability of the second conversion behavior, and the resource usage duration.

As an exemplary embodiment, fig. 5 is a flowchart illustrating a method for predicting an occurrence probability of a first transition behavior of a user according to the present disclosure. Referring to fig. 5, the method for predicting the occurrence probability of the first transition behavior of the user provided by the present disclosure includes the following steps.

In step S501, the user characteristics and the resource characteristics are input into the third user conversion behavior prediction model.

In step S502, a first conversion behavior occurrence probability is determined based on an output of the third user conversion behavior prediction model.

In the disclosure, the third user conversion behavior prediction model is obtained by pre-training based on third sample data, where the third sample data includes user characteristics, resource characteristics, a first label, a second label, and resource usage duration information. The first label is used for indicating that the user carries out first conversion action on the resource corresponding to the resource characteristic. And the second label is used for indicating that the user carries out second conversion action on the resource corresponding to the resource characteristic.

In an exemplary embodiment of the present disclosure, the resource usage duration information may be a duration of a page browsed by a user. The third user transformation behavior prediction model which is trained in advance through the user resource use duration information is more accurate in probability prediction of the transformation behavior of the user.

In an embodiment of the disclosure, the output of the third user conversion behavior prediction model includes a first conversion behavior occurrence probability and a second conversion behavior occurrence probability. Therefore, in the embodiment of the present disclosure, resource recommendation may be performed based on the first conversion behavior occurrence probability and the second conversion behavior occurrence probability.

In the embodiment of the disclosure, resource recommendation is performed based on the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior, the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior are fused, resource recommendation is performed with respect to the second conversion behavior with a sparse dependence number, the accuracy of resource recommendation can be improved, and the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior are predicted by a third user conversion behavior prediction model which is trained in advance based on the first label, the second label and the resource use duration information, so that the prediction accuracy is higher.

As an exemplary embodiment, fig. 6 is a flowchart illustrating a method for predicting occurrence probability of a user conversion behavior according to the method for predicting a user conversion behavior provided by the present disclosure. Referring to fig. 6, the method for predicting user transformation behavior provided by the present disclosure includes the following steps.

In step S601, a first conversion behavior occurrence probability and a second conversion behavior occurrence probability are acquired.

The first conversion behavior occurrence probability and the second conversion behavior occurrence probability may be obtained based on the second user conversion behavior prediction model, or may be obtained based on the third user conversion behavior prediction model.

In step S602, a weighted calculation is performed on the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior, so as to obtain a target probability of the occurrence of the conversion behavior of the target user.

In the present disclosure, the target probability of the target user transition behavior is obtained by performing weighted calculation according to the first transition behavior occurrence probability and the second transition behavior occurrence probability, for example, the weight of the first transition behavior occurrence probability is a, the weight of the second transition behavior occurrence probability is b, and the target probability of the target user transition behavior is (a ^ first transition behavior occurrence probability) + (b ^ second transition behavior occurrence probability).

It can be understood that the occurrence probability of the first conversion behavior cannot replace the occurrence probability of the second conversion behavior all the time, but a certain relationship exists between the two, and the second conversion behavior may be triggered only after the first conversion behavior occurs. According to the target probability obtained by the calculation, the similarity between the target probability and the user characteristics is higher, and the resource characteristics and the user characteristics can be combined more accurately.

The following describes a process of resource recommendation by applying the resource recommendation method according to the embodiment of the present disclosure. In the embodiment of the present disclosure, a Feed platform is used to perform resource recommendation for multiple presentations.

In an exemplary embodiment, when a user uses a Feed platform, the platform acquires user characteristics and multivariate emerging resource characteristics, and obtains a first conversion behavior occurrence probability of the user according to a first user conversion behavior prediction model. And predicting judgment of the conversion behavior of the user based on the first conversion behavior occurrence probability, and recommending the multi-element change resource to the user based on a judgment result.

In another exemplary embodiment, when a user uses a Feed platform, the platform acquires data according to the click behavior of the user and the conversion of the user, and predicts the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior. And performing weighted calculation on the occurrence probability of the first user conversion behavior and the occurrence probability of the second user conversion behavior, namely (a ^ first conversion behavior occurrence probability) + (b ^ second conversion behavior occurrence probability), so as to obtain the target probability of the target user conversion behavior. And recommending the multi-element showing resource to the user based on the target probability of the target user conversion behavior.

In still another exemplary embodiment, when the user uses the Feed platform, the platform acquires the user click behavior, data that the user has converted and the user resource usage duration information, and predicts the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior. And performing weighted calculation on the occurrence probability of the first user conversion behavior and the occurrence probability of the second user conversion behavior, namely (a ^ first conversion behavior occurrence probability) + (b ^ second conversion behavior occurrence probability), so as to obtain the target probability of the target user conversion behavior. And recommending the multi-element showing resource to the user based on the target probability of the target user conversion behavior.

When the multivariate emergence resource is recommended to the user based on the target probability of the target user transformation behavior, for example, the emergence resource can be pushed to the user with the higher probability value of the target user transformation behavior. The multivariate achievement resources obtained by each user through pushing are different, so that the resources which are most interested in and have the highest conversion rate can be recommended for the users, and the conversion rate of the multivariate achievement resources is higher.

Based on the same concept, the embodiment of the present disclosure may use the first user transformation behavior as a training sample, enrich the second user transformation behavior, and perform training of the user transformation behavior prediction model.

In an exemplary embodiment, under the condition that the transformation behaviors are sparse, the first transformation behavior is used for simulating the second transformation behavior to perform model training, so that the second transformation behavior is enriched, and the prediction accuracy of the user transformation behavior prediction model is improved.

In one implementation, in this embodiment of the disclosure, training of the first user conversion behavior prediction model may be performed based on the first conversion behavior.

As an exemplary embodiment, fig. 7 is a schematic flow chart of training a first user conversion behavior prediction model according to the present disclosure. Referring to fig. 7, the method for training the first user transformation behavior prediction model provided by the present disclosure includes the following steps.

In step S701, first sample data is acquired.

In the disclosure, the first sample data includes a user characteristic, a resource characteristic, and a first label, where the first label is used to indicate that the user performs a first conversion action on a resource corresponding to the resource characteristic, the first conversion action is associated with a second conversion action, and an output of the first user conversion action prediction model is a first conversion action occurrence probability for a previous operation action in which the second conversion action occurs.

In the present disclosure, the first sample data is determined according to data that matches the first sample data in the user's historical conversion behavior.

In step S702, the user features and the resource features in the first sample data are input to the first prediction model as input features, so as to obtain a predicted value of the occurrence probability of the first conversion behavior.

In step S703, the first prediction model is trained based on the first label and the predicted value of the occurrence probability of the first conversion behavior until the first prediction model converges, so as to obtain a first user conversion behavior prediction model.

In the embodiment of the present disclosure, the input of the first user conversion behavior prediction model is a user characteristic and a resource characteristic, and the output of the first user conversion behavior prediction model is a probability of occurrence of the first user conversion behavior.

The training process of the first user conversion behavior prediction model may adopt a conventional training process, and the difference is that the first sample data serving as the training sample includes a first label for indicating that the user performs a first conversion behavior on the resource corresponding to the resource feature. The user conversion behavior prediction method based on the multi-source transformation behavior comprises the steps that training is conducted on the basis of user characteristics, resource characteristics and a first label for identifying a first conversion behavior, and the number of second conversion behaviors can be enriched through the first conversion behavior, so that the prediction accuracy is higher compared with a user conversion behavior prediction model obtained through training based on the second conversion behaviors, and therefore the probability prediction of the first conversion behavior is more accurate.

In the embodiment of the disclosure, under the condition that the second conversion behavior is sparse, the training of the user conversion model used for resource recommendation can be performed in a manner of training the first user conversion behavior, so as to alleviate the problem of sparse conversion behavior.

However, as the second conversion behavior is gradually accumulated, the number of samples of the second conversion behavior is also increased, for example, the third-party platform may pass the second conversion behavior back, and the problem of sparseness of samples of the second conversion behavior is alleviated, so in the embodiment of the present disclosure, the first conversion behavior and the second conversion behavior may be combined to perform training of the user conversion behavior model, that is, training of the second user conversion behavior model may be performed.

As an exemplary embodiment, fig. 8 is a schematic flow chart of training a second user conversion behavior prediction model according to the present disclosure. Referring to fig. 8, the method for training the second user transformation behavior prediction model provided by the present disclosure includes the following steps.

In step S801, second sample data is determined.

In the present disclosure, the second sample data includes a user characteristic, a resource characteristic, a first tag, and a second tag, where the first tag is used to indicate that the user performs a first conversion action on a resource corresponding to the resource characteristic, the second tag is used to indicate that the user performs a second conversion action on a resource corresponding to the resource characteristic, an input of the second user conversion action prediction model is the user characteristic and the resource characteristic, and an output is an occurrence probability of the first conversion action and an occurrence probability of the second conversion action.

In step S802, the user characteristics and the resource characteristics in the second sample data are input to the second prediction model, so as to obtain a predicted value of the occurrence probability of the first conversion behavior and a predicted value of the occurrence probability of the second conversion behavior.

In step S803, a second prediction model is trained based on the first conversion behavior occurrence probability predicted value and the first label, and based on the second conversion behavior occurrence probability predicted value and the second label, so as to obtain a second user conversion behavior prediction model.

In the embodiment of the present disclosure, the input of the second user conversion behavior prediction model is a user characteristic and a resource characteristic, and the output of the second user conversion behavior prediction model is a probability of occurrence of the first user conversion behavior and a probability of occurrence of the second user conversion behavior.

The training process of the second user conversion behavior prediction model may adopt a conventional training process, and the difference is that the training samples include second sample data. The second sample data comprises a first label used for indicating that the user carries out a first conversion action on the resource corresponding to the resource feature and a second label used for indicating a second conversion action. In addition, the training process of the occurrence probability of the first user transformation behavior can be used as an auxiliary task for training the prediction of the occurrence probability of the second user transformation behavior, and the prediction accuracy of the occurrence probability of the second user transformation behavior is improved.

Fig. 9 is a schematic diagram of a training architecture of a second user translation behavior prediction model provided in accordance with the present disclosure. Referring to fig. 9, the input of the second user transformation behavior prediction model includes a user characteristic, a resource characteristic, a first label indicating the first user transformation behavior, and a second label indicating the second user transformation behavior, and the output includes the occurrence probability of the first transformation behavior and the occurrence probability of the second transformation behavior, so that the joint training of the occurrence probability of the first transformation behavior and the occurrence probability of the second transformation behavior is realized, and the prediction accuracy of the occurrence probability of the transformation behavior is improved.

And training is carried out based on the user characteristics, the resource characteristics and the first label for identifying the first conversion behavior, and the number of the second conversion behaviors can be enriched by the first conversion behavior, so that the prediction accuracy is higher compared with a user conversion behavior prediction model obtained by training based on the second conversion behavior, and the probability prediction of the first conversion behavior is more accurate.

According to the second user conversion behavior prediction model provided by the disclosure, the actual conversion rate of the user, the occurrence probability of the first conversion behavior and the model prediction precision are all improved.

Furthermore, after the joint prediction of the first transformation behavior and the second transformation behavior is performed, in order to more accurately improve the prediction accuracy of the transformation behavior, the resource usage duration may be further introduced as a training target.

Fig. 10 is a schematic diagram illustrating a correspondence relationship between a resource usage duration and a user conversion behavior occurrence probability. Referring to fig. 10, as the resource usage duration increases, the probability of occurrence of the user conversion behavior also increases, and as the user conversion rate increases, the user tends to increase the resource usage duration, that is, the user conversion behavior and the user have a positive relationship with the resource usage duration. Therefore, the resource use duration is introduced to train the user transformation behavior prediction model, and the prediction accuracy of the user transformation behavior prediction model can be further improved. The user conversion behavior prediction model trained by introducing the resource use duration can be understood as a third user conversion behavior prediction model.

As an exemplary embodiment, fig. 11 is a schematic flow chart of training a third user conversion behavior prediction model provided according to the present disclosure. Referring to fig. 11, the method for training the third user transformation behavior prediction model provided by the present disclosure includes the following steps.

In step S1101, third sample data is determined.

In this disclosure, the third sample data includes a user characteristic, a resource characteristic, a first tag, a second tag, and resource use duration information, where the first tag is used to indicate that the user has performed a first conversion behavior on the resource corresponding to the resource characteristic, and the second tag is used to indicate that the user has performed a second conversion behavior on the resource corresponding to the resource characteristic.

In step S1102, the user characteristics and the resource characteristics in the third sample data are input to the third prediction model, so as to obtain a predicted value of the occurrence probability of the first conversion behavior and a predicted value of the occurrence probability of the second conversion behavior.

In step S1103, a third prediction model is trained based on the first conversion behavior occurrence probability prediction value and the first label, the second conversion behavior occurrence probability prediction value and the second label, and the resource usage duration information, so as to obtain a third user conversion behavior prediction model.

The training process of the third user conversion behavior prediction model may adopt a conventional training process, and the difference is that the training samples include third sample data. The second sample data comprises a first label used for indicating that a user carries out a first conversion behavior on the resource corresponding to the resource feature, a second label used for indicating a second conversion behavior and resource use duration information. In addition, the training process of the occurrence probability of the first user transformation behavior and the training process of the resource use duration can be used as an auxiliary task for training the prediction of the occurrence probability of the second user transformation behavior, a mode of user transformation behavior model training of a multi-objective task is achieved, and the prediction accuracy of the occurrence probability of the second user transformation behavior is improved.

FIG. 12 illustrates a network architecture diagram of a third user translation behavior prediction model in an exemplary embodiment of the present disclosure. Referring to fig. 12, the input of the third user conversion behavior prediction model includes a user characteristic, a resource characteristic, a first label indicating a first user conversion behavior, and a second label indicating a second user conversion behavior, and the output includes an occurrence probability of the first conversion behavior, an occurrence probability of the second conversion behavior, and a resource usage duration, so that joint training of the occurrence probability of the first conversion behavior, the occurrence probability of the second conversion behavior, and the resource usage duration is achieved.

In an exemplary embodiment, the input of the third user conversion behavior prediction model may be four input values of user click behavior, user request, user resource usage duration and conversion, and the third user conversion behavior prediction model is trained.

The user click behavior and the user request belong to the first conversion behavior, and the user has more resource use time and sample size of the first conversion behavior, and the behavior of the user is similar to that of the second conversion behavior, so that the third user conversion behavior prediction model can learn the user behavior characteristics from the user click behavior and the user request, and further more accurate prediction is made on the second conversion behavior.

In the method for predicting the user conversion behavior provided by the embodiment of the disclosure, the previous operation behavior related to the conversion behavior is brought into the user conversion behavior prediction model for training based on the relationship between the operation behavior of the previous operation of the conversion behavior and the target conversion behavior.

In the training process of the three user transformation behavior prediction models provided by the embodiment of the disclosure, a shared hidden network exists, and the shared hidden network can play a role in updating any one of the first user transformation behavior prediction model, the second user transformation behavior prediction model and the third user behavior transformation prediction model, so that the results of the other user transformation behavior prediction models can be positively influenced, and the prediction results of the user transformation behavior prediction models are more accurate.

In the embodiment of the disclosure, in the training process of the user conversion behavior prediction model, the user behavior characteristics can be learned from other behaviors related to conversion, and the prediction accuracy of the user conversion behavior prediction model is improved. Furthermore, the training method for the user transformation behavior prediction model provided by the embodiment of the disclosure can improve the model training effect without changing the model updating period, so that the estimation effect of the model and other users can be further improved on the basis of ensuring the estimation effect of the cold start user.

Based on the same conception, the embodiment of the disclosure also provides a device for predicting the user conversion behavior.

It is understood that, in order to implement the above functions, the apparatus for predicting user conversion behavior provided by the embodiments of the present disclosure includes a hardware structure and/or a software module for performing each function. The disclosed embodiments can be implemented in hardware or a combination of hardware and computer software, in combination with the exemplary elements and algorithm steps disclosed in the disclosed embodiments. Whether a function is performed as hardware or computer software drives hardware depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

As an exemplary embodiment, fig. 13 is a block diagram of an apparatus for predicting user conversion behavior according to an exemplary embodiment shown in the present disclosure. Referring to fig. 13, the apparatus 1300 for predicting user conversion behavior provided by the present disclosure includes an obtaining unit 1301 and a predicting unit 1302.

The obtaining unit 1301 is configured to obtain a user characteristic and a resource characteristic. The predicting unit 1302 is configured to predict an occurrence probability of a first transformation behavior according to the user characteristics and the resource characteristics, where the occurrence probability of the first transformation behavior is a probability of a user with the user characteristics performing the first transformation behavior on the resource corresponding to the resource characteristics, the first transformation behavior is associated with a second transformation behavior and is a previous operation behavior in which the second transformation behavior occurs, and the second transformation behavior is a preset target user transformation behavior that needs to be predicted to occur.

As an exemplary embodiment, the prediction unit 1302 predicts the occurrence probability of the first transition behavior in the following manner: the user characteristics and the resource characteristics are input into a first user conversion behavior prediction model, the occurrence probability of the first conversion behavior is determined based on the output of the first user conversion behavior prediction model, the first user conversion behavior prediction model is obtained through pre-training based on first sample data, the first sample data comprises the user characteristics, the resource characteristics and a first label, the first label is used for indicating that a user carries out first conversion behavior on resources corresponding to the resource characteristics, the input of the first user conversion behavior prediction model is the user characteristics and the resource characteristics, and the output is the occurrence probability of the first conversion behavior.

The prediction unit 1302 is configured to predict the occurrence probability of the second conversion behavior in the following manner: inputting the user characteristics and the resource characteristics into a second user conversion behavior prediction model, and determining the occurrence probability of a first conversion behavior based on the output of the second user conversion behavior prediction model; the second user transformation behavior prediction model is obtained by pre-training based on second sample data, the second sample data comprises user characteristics, resource characteristics, a first label and a second label, the first label is used for indicating that a user carries out a first transformation behavior on resources corresponding to the resource characteristics, the second label is used for indicating that the user carries out a second transformation behavior on resources corresponding to the resource characteristics, and the input of the second user transformation behavior prediction model is the user characteristics and the resource characteristics and the output is the occurrence probability of the first transformation behavior and the occurrence probability of the second transformation behavior.

The prediction unit 1302 is configured to predict the occurrence probability of the first transition behavior in the following manner: inputting the user characteristics and the resource characteristics into a third user conversion behavior prediction model, determining the occurrence probability of a first conversion behavior based on the output of the third user conversion behavior prediction model, pre-training the third user conversion behavior prediction model based on third sample data, wherein the third sample data comprises the user characteristics, the resource characteristics, a first label, a second label and resource use duration information, the first label is used for indicating that a user performs a first conversion behavior on resources corresponding to the resource characteristics, the second label is used for indicating that the user performs a second conversion behavior on resources corresponding to the resource characteristics, the input of the third user conversion behavior prediction model is the user characteristics and the resource characteristics, and the output is the occurrence probability of the first conversion behavior, the occurrence probability of the second conversion behavior and the occurrence probability of the resource use duration conversion behavior.

For the obtaining unit 1301, any one of the devices in the predicting unit 1302 is further configured to: and weighting the occurrence probability of the first conversion behavior and the occurrence probability of the second conversion behavior to obtain the target probability of the target user conversion behavior.

As an exemplary embodiment, fig. 14 is a block diagram of an apparatus for training a user transformation behavior prediction model according to the present disclosure. Referring to fig. 14, the predictive model apparatus 1400 for training user conversion behavior provided by the present disclosure includes an obtaining unit 1401, a predicting unit 1402, and a training unit 1403.

The obtaining unit 1401 is configured to obtain first sample data, where the first sample data includes a user characteristic, a resource characteristic, and a first tag, and the first tag is used to indicate that a user performs a first conversion action on a resource corresponding to the resource characteristic, and the first conversion action is associated with a second conversion action and is a previous operation action in which the second conversion action occurs;

a prediction unit 1402, configured to input the user characteristic and the resource characteristic in the first sample data as input characteristics to a first prediction model, so as to obtain a predicted value of a first conversion behavior occurrence probability;

a training unit 1403, configured to train the first prediction model according to the first label and the first conversion behavior occurrence probability prediction value until the first prediction model converges to obtain a first user conversion behavior prediction model, where an output of the first user conversion behavior prediction model is the first conversion behavior occurrence probability.

As an exemplary embodiment, fig. 15 is a block diagram of an apparatus for training a user translation behavior prediction model according to the present disclosure. Referring to fig. 15, the predictive model apparatus 1500 for training the user conversion behavior provided by the present disclosure includes a determination unit 1501, a prediction unit 1502, and a training unit 1503.

The determining unit 1501 is configured to determine second sample data, where the second sample data includes a user characteristic, a resource characteristic, a first tag and a second tag, the first tag is used to indicate that a user performs a first conversion behavior on a resource corresponding to the resource characteristic, and the second tag is used to indicate that the user performs a second conversion behavior on a resource corresponding to the resource characteristic;

the prediction unit 1502 is configured to input the user characteristics and the resource characteristics in the second sample data as input characteristics to the second prediction model, so as to obtain a first conversion behavior occurrence probability prediction value and a second conversion behavior occurrence probability prediction value;

the training unit 1503 is configured to train the second prediction model according to the first conversion behavior occurrence probability prediction value and the first label, and based on the second conversion behavior occurrence probability prediction value and the second label, to obtain a second user conversion behavior prediction model, where the second user conversion behavior prediction model is output as the first conversion behavior occurrence probability and the second conversion behavior occurrence probability.

As an exemplary implementation, fig. 16 is a block diagram of a device for predicting a user transformation behavior according to the present disclosure, and referring to fig. 16, a device 1600 for predicting a user transformation behavior according to the present disclosure includes a determining unit 1601, a predicting unit 1602, and a training unit 1603.

The determining unit 1601 is configured to determine third sample data, where the third sample data includes a user characteristic, a resource characteristic, a first tag, a second tag, and resource usage duration information, the first tag is used to indicate that a user performs a first conversion behavior on a resource corresponding to the resource characteristic, and the second tag is used to indicate that the user performs a second conversion behavior on the resource corresponding to the resource characteristic;

a prediction unit 1602, configured to input the user characteristics and the resource characteristics in the third sample data as input characteristics to a third prediction model, so as to obtain a first conversion behavior occurrence probability prediction value and a second conversion behavior occurrence probability prediction value;

the training unit 1603 is configured to train the third prediction model according to the first conversion behavior occurrence probability prediction value and the first label, and based on the second conversion behavior occurrence probability prediction value and the second label, and the resource duration prediction value and the resource use duration information, to obtain a third user conversion behavior prediction model, where the output of the third user conversion behavior prediction model is the first conversion behavior occurrence probability, the second conversion behavior occurrence probability, and the resource use duration.

The specific manner in which the various modules perform operations has been described in detail in relation to the apparatus of the present disclosure above, and will not be elaborated upon here.

In the technical scheme of the disclosure, the acquisition, storage, application and the like of the personal information of the related user all accord with the regulations of related laws and regulations, and do not violate the good customs of the public order.

The present disclosure also provides an electronic device, a readable storage medium, and a computer program product according to embodiments of the present disclosure.

Fig. 17 illustrates a schematic block diagram of an example electronic device 1700 that can be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 17, the apparatus 1700 includes a computing unit 1701 that may perform various appropriate actions and processes in accordance with a computer program stored in a Read Only Memory (ROM)1702 or a computer program loaded from a storage unit 1708 into a Random Access Memory (RAM) 1703. In the RAM 1703, various programs and data required for the operation of the device 1700 can also be stored. The computing unit 1701, the ROM 1702, and the RAM 1703 are connected to each other through a bus 1704. An input/output (I/O) interface 1705 is also connected to bus 1704.

Various components in the device 1700 are connected to the I/O interface 1705, including: an input unit 1706 such as a keyboard, a mouse, and the like; an output unit 1707 such as various types of displays, speakers, and the like; a storage unit 1708 such as a magnetic disk, optical disk, or the like; and a communication unit 1709 such as a network card, modem, wireless communication transceiver, etc. The communication unit 1709 allows the device 1700 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The computing unit 1701 may be a variety of general purpose and/or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 1701 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various dedicated Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, and the like. The computing unit 1701 performs various methods and processes described above, such as methods of predicting user translation behavior and/or methods of training models. For example, in some embodiments, the methods of predicting user translation behavior and/or the methods of training the models may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1708. In some embodiments, part or all of a computer program may be loaded and/or installed onto device 1700 via ROM 1702 and/or communications unit 1709. When loaded into RAM 1703 and executed by computing unit 1701, may perform one or more of the steps of the above-described method of predicting user translation behavior and/or method of training a model. Alternatively, in other embodiments, the computing unit 1701 may be configured in any other suitable manner (e.g., via firmware) to perform methods of predicting user translation behavior and/or methods of training models.

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuitry, Field Programmable Gate Arrays (FPGAs), Application Specific Integrated Circuits (ASICs), Application Specific Standard Products (ASSPs), system on a chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, receiving data and instructions from, and transmitting data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for implementing the methods of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a back-end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), Wide Area Networks (WANs), and the Internet.

The computer system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server with a combined blockchain.

It should be understood that various forms of the flows shown above may be used, with steps reordered, added, or deleted. For example, the steps described in the present disclosure may be executed in parallel, sequentially, or in different orders, as long as the desired results of the technical solutions disclosed in the present disclosure can be achieved, and the present disclosure is not limited herein.

The above detailed description should not be construed as limiting the scope of the disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may be made in accordance with design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present disclosure should be included in the scope of protection of the present disclosure.

Claims

1. A method of predicting user conversion behavior, comprising:

acquiring user characteristics and resource characteristics;

2. The method of claim 1, predicting a first transition behavior occurrence probability based on the user characteristics and the resource characteristics, comprising:

3. The method of claim 1, predicting a first transition behavior occurrence probability based on the user characteristics and the resource characteristics, comprising:

the second sample data comprises user characteristics, resource characteristics, a first label and a second label, the first label is used for indicating that a user performs a first conversion behavior on the resource corresponding to the resource characteristics, and the second label is used for indicating that the user performs a second conversion behavior on the resource corresponding to the resource characteristics;

4. The method of claim 1, wherein predicting a first transition behavior occurrence probability based on the user characteristics and the resource characteristics comprises:

5. The method of claim 3 or 4, further comprising:

6. A method of training a user translation behavior prediction model, comprising:

obtaining first sample data, wherein the first sample data comprises user characteristics, resource characteristics and a first label, the first label is used for indicating that a user carries out a first conversion behavior on a resource corresponding to the resource characteristics, the first conversion behavior is associated with a second conversion behavior, and the first conversion behavior is a previous operation behavior of the second conversion behavior;

7. A method of training a user translation behavior prediction model, comprising:

the output of the second user conversion behavior prediction model is a first conversion behavior occurrence probability and a second conversion behavior occurrence probability.

8. A method of training a user translation behavior prediction model, comprising:

9. An apparatus for predicting user conversion behavior, comprising:

10. The apparatus of claim 9, wherein the prediction unit is configured to predict the probability of occurrence of the first transition behavior based on the user characteristic and the resource characteristic in the following manner:

11. The apparatus of claim 9, wherein the prediction unit is configured to predict the probability of occurrence of the first transition behavior based on the user characteristic and the resource characteristic in the following manner:

12. The apparatus of claim 9, wherein the prediction unit is configured to predict the probability of occurrence of the first transition behavior based on the user characteristic and the resource characteristic in the following manner:

13. The apparatus of any of claims 11 to 12, the prediction unit further to:

14. A predictive model apparatus for training user conversion behavior, comprising:

15. An apparatus for training a predictive model of user translation behavior, comprising:

16. A trained user conversion behavior prediction apparatus comprising:

the prediction unit is used for inputting the user characteristics and the resource characteristics in the third sample data into a third prediction model to obtain a first conversion behavior occurrence probability prediction value, a second conversion behavior occurrence probability prediction value and a resource duration prediction value;

and the training unit is used for training the third prediction model according to the first conversion behavior occurrence probability prediction value, the first label, the second conversion behavior occurrence probability prediction value, the second label, the resource duration prediction value and the resource use duration information to obtain a third user conversion behavior prediction model, and the output of the third user conversion behavior prediction model is the first conversion behavior occurrence probability, the second conversion behavior occurrence probability and the resource use duration.

17. An electronic device, comprising:

at least one processor; and

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

18. A non-transitory computer readable storage medium having stored thereon computer instructions for causing the computer to perform the method of any one of claims 1-8.

19. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1-8.