CN107993085B

CN107993085B - Model training method, and user behavior prediction method and device based on model

Info

Publication number: CN107993085B
Application number: CN201710980102.8A
Authority: CN
Inventors: 马书超
Original assignee: Advanced New Technologies Co Ltd
Current assignee: Advanced Nova Technology Singapore Holdings Ltd
Priority date: 2017-10-19
Filing date: 2017-10-19
Publication date: 2021-05-18
Anticipated expiration: 2037-10-19
Also published as: CN113409084A; CN107993085A

Abstract

The embodiment of the specification can train a prediction model for predicting the specific behavior of a user according to a training sample and a specific algorithm, when the specific behavior of the user needs to be predicted, data related to the specific behavior of the user are obtained, features related to the specific behavior are extracted from the data related to the specific behavior, the features related to the specific behavior are input into the prediction model to obtain an output value, and the specific behavior of the user is predicted according to the output value. Because the data of the user related to the specific behavior can reflect the behavior intention of the user to a great extent, the embodiment of the specification can more accurately predict the specific behavior of the user, and then push the related information according to the predicted specific behavior, thereby improving the accuracy of pushing the information.

Description

Model training method, and user behavior prediction method and device based on model

Technical Field

The application relates to the technical field of computers, in particular to a model training method, a user behavior prediction method based on a model and a user behavior prediction device based on the model.

Background

With the rapid development of the internet technology, a service platform based on the internet technology brings more and more convenience to people, for example, a tour application can push some tour information to people. If the specific behavior of the user can be predicted, the pushed information is definitely more accurate according to the pushing of the related information by the specific behavior.

Disclosure of Invention

An object of the embodiments of the present disclosure is to provide a model training method, a user behavior prediction method based on a model, and an apparatus thereof, so as to predict a specific behavior of a user more accurately, and then push related information according to the predicted specific behavior, thereby improving the accuracy of pushing information.

In order to solve the above technical problem, the embodiments of the present specification are implemented as follows:

in a first aspect, a model training method is provided, the method including:

acquiring a training sample set, wherein the training sample set comprises training samples for training a model, and the training samples comprise data of a user, which are related to specific behaviors;

extracting features of the training samples in the training sample set, which are related to specific behaviors;

and training the characteristics related to the specific behaviors according to a specific algorithm to obtain a prediction model, wherein the prediction model is used for establishing a mapping relation between the characteristics related to the specific behaviors and the specific behavior prediction information of the user.

In a second aspect, a model-based user behavior prediction method is provided, the method comprising:

acquiring data related to specific behaviors of a target user;

extracting features related to the specific behaviors from the data related to the specific behaviors;

inputting the characteristics related to the specific behaviors into a prediction model to obtain an output value, wherein the prediction model is obtained by training the characteristics related to the specific behaviors of a training sample by using a specific algorithm, and the prediction model is used for establishing a mapping relation between the characteristics related to the specific behaviors and the specific behavior prediction information of a user;

and predicting the specific behavior of the target user according to the output value.

In a third aspect, there is provided a model training apparatus, the apparatus comprising:

a first obtaining unit, configured to obtain a training sample set, where the training sample set includes training samples for training a model, and the training samples include data of a user related to a specific behavior;

the first extraction unit is used for extracting the characteristics, related to the specific behaviors, of the training samples in the training sample set acquired by the first acquisition unit;

and the training unit is used for training the features which are extracted by the first extraction unit and are related to the specific behaviors according to a specific algorithm to obtain a prediction model, and the prediction model is used for establishing a mapping relation between the features which are related to the specific behaviors and the prediction information of the specific behaviors of the user.

In a fourth aspect, there is provided a model-based user behavior prediction apparatus, the apparatus comprising:

a second acquisition unit configured to acquire data of a target user relating to a specific behavior;

a second extracting unit, configured to extract a feature related to the specific behavior from the data related to the specific behavior acquired by the second acquiring unit;

the processing unit is used for inputting the features related to the specific behaviors extracted by the second extraction unit into a prediction model to obtain an output value, the prediction model is obtained by training the features related to the specific behaviors of the training sample by using a specific algorithm, and the prediction model is used for establishing a mapping relation between the features related to the specific behaviors and the prediction information of the specific behaviors of the user;

and the prediction unit is used for predicting the specific behavior of the target user according to the output value obtained by the processing unit.

In a fifth aspect, an electronic device is provided, comprising:

a processor; and

a memory arranged to store computer executable instructions that, when executed, cause the processor to:

In a sixth aspect, an electronic device is provided, comprising:

a processor; and

acquiring data related to specific behaviors of a target user;

In a seventh aspect, a computer storage medium is provided that stores one or more programs that, when executed by an electronic device that includes a plurality of application programs, cause the electronic device to:

In an eighth aspect, a computer storage medium is provided that stores one or more programs that, when executed by an electronic device that includes a plurality of application programs, cause the electronic device to:

acquiring data related to specific behaviors of a target user;

As can be seen from the above technical solutions provided by the embodiments of the present specification, in the embodiments of the present specification, a prediction model for predicting a specific behavior of a user may be trained according to a training sample and a specific algorithm, when the specific behavior of the user needs to be predicted, data related to the specific behavior of the user is obtained, features related to the specific behavior are extracted from the data related to the specific behavior, the features related to the specific behavior are input into the prediction model to obtain an output value, and the specific behavior of the user is predicted according to the output value. Because the data of the user related to the specific behavior can reflect the behavior intention of the user to a great extent, the embodiment of the specification can more accurately predict the specific behavior of the user, and then push the related information according to the predicted specific behavior, thereby improving the accuracy of pushing the information.

Drawings

In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort.

FIG. 1 is a flow diagram of a model training method according to one embodiment of the present description;

FIG. 2 is a flow diagram of a model-based user behavior prediction method according to one embodiment of the present description;

FIG. 3 is a diagram of an example of a model-based user behavior prediction method according to one embodiment of the present description;

FIG. 4 is a diagram of an example of a model-based user behavior prediction method according to another embodiment of the present description;

FIG. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present description;

fig. 6 is a schematic structural diagram of an electronic device according to another embodiment of the present specification;

FIG. 7 is a schematic diagram of a model training apparatus according to an embodiment of the present disclosure;

fig. 8 is a schematic structural diagram of a model-based user behavior prediction apparatus according to an embodiment of the present disclosure.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort shall fall within the protection scope of the present specification.

The embodiment of the specification provides a model training method, a user behavior prediction method based on a model and a user behavior prediction device based on the model.

First, a model training method provided in the embodiments of the present specification is described below.

FIG. 1 is a flow diagram of a model training method according to an embodiment of the present disclosure, which may include the following steps, as shown in FIG. 1:

in S102, a training sample set including training samples for training a model is obtained, the training samples including data of a user related to a specific behavior.

In the embodiment of the present specification, the specific behavior may include: a tour, purchase, or fitness activity, and the like. For ease of understanding, the following description will mainly describe the embodiments of the present specification with reference to the tour behavior as an example, and other behaviors are similar to the tour behavior.

In an embodiment of the present specification, the training sample set includes a plurality of training samples, where the plurality of training samples includes: a plurality of positive samples and a plurality of negative samples, typically, the positive samples include: data relating to a particular activity for a user having the particular activity, the negative examples comprising: there has never been data relating to a particular behavior for users who have experienced that particular behavior. When the specific behavior is an outbound behavior, the positive samples include: data relating to the tour for users who have taken a tour, the negative examples include: data relating to the outbound of a user who never has taken an outbound action.

In embodiments of the present description, the data relating to a particular action may include any combination of one or more of the following: search records, purchase records, collection records, browsing records, and application click times, and the data related to a particular behavior may further include, in any combination: user basic information and a user tag. The application program may be all application programs installed in the terminal device of the user, for example, all application programs installed in an Android operating system mobile phone, or all application programs in one application, for example, all application programs in a "pay for treasure application", and the user tag is information used for characterizing whether the user is a specific behavior enthusiast.

When the specific behavior is an outbound behavior, in one example, the positioning information (the position information of the outbound place) of the user who has the outbound behavior, the outbound time T, and the data related to the outbound within 30 days before the time T may be determined as positive samples.

In S104, features of the training samples in the training sample set related to the specific behavior are extracted.

Optionally, in one embodiment, the data related to the specific behavior of the training sample in S102 includes: searching for a record, in this case, the S104 may include:

extracting keywords related to a specific behavior from the search records;

the keyword is taken as one of the feature related to the specific behavior or the feature related to the specific behavior.

When the specific behavior is a tour behavior, in one example, the search record records search time and search sequence, and the search sequence in the search record can be cut into words, and then keywords related to tour are extracted, and the keywords themselves are used as features related to tour. Or, some special words can be processed to form a combined feature, and the combined feature is taken as a feature related to the outing, wherein the combined feature includes two types:

in the first category, the national vocabulary in the search sequence is specially processed, and for such words, the vocabulary is manually established.

And secondly, performing special processing on words such as travel words, wifi words and the like in the search sequence, and manually establishing a word list for the words, wherein the specific processing process is as follows: judging whether the user searches for words related to the country in the near period of time, if so, combining the words and the country together to serve as characteristics; for example, a user has searched for "travel," if the user has searched for "Thailand" in the near future, Thailand + travel is used as the combined feature, and if only travel is searched, the word is not used as the feature and is filtered out directly.

Optionally, in another embodiment, the data related to the specific behavior of the training sample in S102 includes: the purchase record, at this time, the S104 may include:

extracting an article number and an article type related to a specific behavior from the purchase record;

the item number and the item category are taken as one of the characteristics related to the specific behavior or the characteristics related to the specific behavior.

When the specific behavior is an outbound behavior, in an example, the purchase record includes the purchase time and the related information of the purchased article, and the article number id and the article type category may be extracted from the related information of the purchased article, where the article may include: real items and virtual items, wherein the virtual items may be outbound orders or the like.

Optionally, in another embodiment, the data related to the specific behavior of the training sample in S102 includes: the collection record, in this case, the S104 may include:

extracting the item numbers and the item types related to the specific behaviors from the collection records;

Optionally, in another embodiment, the data related to the specific behavior of the training sample in S102 includes: browsing the records, in this case, the S104 may include:

extracting the article number and the article type related to the specific behavior from the browsing record;

It should be noted that, since the feature extraction processes of the collection record and the browsing record are similar to the feature extraction process of the purchase record, the detailed description is omitted here.

Optionally, in another embodiment, the data related to the specific behavior of the training sample in S102 includes: the number of times of application clicks, in this case, S104 may include:

calculating the ratio of the application program clicking times related to the specific behavior according to the application program clicking times;

the duty value is taken as one of the characteristic related to the specific behavior or the characteristic related to the specific behavior.

When the specific action is the outbound action, in an example, the S104 may include: calculating the ratio of the application program clicking times related to the outing according to the application program clicking times; and taking the proportion value as one of the characteristics related to the outing or the characteristics related to the outing.

For example, the application program apps related to the outbound in the "pay for use application" are screened in advance, and according to the click log of the user on the apps in the "pay for use application", the number of clicks of the user on the apps and the total number of clicks on all apps in the "pay for use application" are obtained, so that a ratio value between [0 and 1] is obtained.

Optionally, in another embodiment, the data related to the specific behavior of the training sample in S102 may further include, on the basis of any of the above embodiments: the system comprises user basic information and a user tag, wherein the user tag is used for representing whether a user is a specific behavior fan; in this case, S104 may include:

extracting basic information characteristics of the user from the basic information of the user, and taking the basic information characteristics of the user and the user label as one of the characteristics related to the specific behavior; wherein, the user basic information characteristics may include: age and occupation, etc.

When the specific behavior is a tour behavior, in one example, the user tag is used to characterize whether the user is a tour fan, and in this case, the user basic information feature and the user tag are taken as one of the features related to the tour.

Preferably, as an example, the data related to the specific behavior of the training sample in S102 includes: the method comprises the steps of searching records, purchasing records, collecting records, browsing records and application program clicking times, at the moment, extracting keywords related to specific behaviors from the searching records of training samples, extracting article numbers and article types related to the specific behaviors from the purchasing records of the training samples, extracting article numbers and article types related to the specific behaviors from the collecting records of the training samples, extracting article numbers and article types related to the specific behaviors from the browsing records of the training samples, calculating the proportion value of the clicking times of the application programs related to the specific behaviors according to the clicking times of the application programs of the training samples, and taking the extracted keywords, article numbers, article types and proportion values as the features related to the specific behaviors of the training samples.

Preferably, as an example, the data related to the specific behavior of the training sample in S102 includes: search records, purchase records, collection records, browsing records, application click times, user basic information, and user tags, at this time, extracting keywords related to a specific behavior from a search record of a training sample, extracting an article number and an article type related to the specific behavior from a purchase record of the training sample, extracting an article number and an article type related to the specific behavior from a collection record of the training sample, extracting an article number and an article type related to the specific behavior from a browsing record of the training sample, calculating a ratio of application program clicks related to the specific behavior according to the application program clicks of the training sample, extracting basic information characteristics of users such as age and occupation from basic information of the users of the training sample, and taking the extracted keywords, article numbers and article types, ratios, basic information characteristics of the users and user labels as the characteristics related to the specific behavior of the training sample.

It should be noted that, because the difference between the positive sample and the negative sample is large, the negative sample generally needs to be sampled, and the specific sampling ratio can be adjusted according to actual needs, which is not limited in the embodiments of the present specification.

In S106, the features of the training samples related to the specific behavior are trained according to a specific algorithm to obtain a prediction model, and the prediction model is used to establish a mapping relationship between the features related to the specific behavior and the specific behavior prediction information of the user.

In the embodiment of the specification, the specific algorithm may comprise a logistic regression algorithm or a neural network algorithm.

Preferably, in the embodiment of the present specification, the specific algorithm is a logistic regression algorithm, and correspondingly, the prediction model is an LR model. Logistic regression is a classification method, mainly used to solve two classification problems (i.e. only two kinds of outputs are output, and each represents two classifications), and the Logistic regression algorithm uses Logistic function (or Sigmoid function), the curve form of the function is S-shaped curve, and the function form is:

for the case of linear boundaries, the boundary form is as follows:

constructing a prediction function using equation (1) and equation (2):

wherein, theta_iIs a weight value, x_iFor the eigenvalues corresponding to the characteristics i, θ^T＝[θ₁,θ₂,...,θ_n]，x＝[x₁,x₂,...,x_n]In general, for a training sample, if the feature i in the training sample satisfies a certain condition, x_iThe value is 1, otherwise the value is 0; the feature value may be other natural numbers, and the present embodiment is not limited to this.

In the embodiment of the present specification, for each training sample, the features of the training sample are quantized into a numerical value, that is, one feature of the training sample corresponds to one numerical value (that is, a characteristic value), the characteristic value of each training sample is vectorized and expressed, and finally, the vectorized and expressed result is substituted into formula (3). A large number of training samples are processed as above to obtain a large number of functions, then the large number of functions are subjected to iterative solution, and theta is calculated^T＝[θ₁,θ₂,...,θ_n]Thus, an LR model, i.e., a probability function consisting of eigenvalues corresponding to a plurality of characteristics and weight values corresponding to each eigenvalue, is obtained:

for the process of quantizing the features of the training samples into numerical values, taking the ratio as an example, the ratio is a value before [0,1], multiplying the ratio by 5, rounding to obtain a number between [0,5], and selecting one number as the feature value.

It should be noted that, in the process of quantizing (converting) the features of the training samples into numerical values, each feature of the training samples can be converted into an appropriate numerical value by using a reasonable rule according to actual needs. In addition, when the LR model is used, the characteristics of the user to be predicted, which are related to the specific behaviors, are input into the LR model, the output of the LR model is a probability value, and the value range of the probability value is 0-1.

In this embodiment of the present specification, LR models for different purposes may be trained by using different features of a training sample, specifically, an LR model for predicting whether a user has an intention to make a specific behavior may be trained, and an LR model for predicting an intention specific behavior of a user may also be trained.

When the specific behavior is an outbound behavior, an LR model for predicting whether the user has an outbound intention may be trained, and an LR model for predicting an outbound destination of the user may also be trained.

Optionally, in one embodiment, when it is desired to train an LR model for predicting whether a user has an intent to make a particular behavior, at least two of the following behavior-specific features of the LR algorithm and training samples may be used: and training the keywords, the item numbers, the item types and the occupation ratios to obtain an LR model.

Preferably, as an example, the LR algorithm is used to train keywords, item numbers and item types, percentage values, user basic information features, and user tags, resulting in an LR model for predicting whether a user has an intention to make a particular behavior.

Alternatively, in another embodiment, the search feature is primarily used when it is desired to train an LR model for predicting the intended specific behavior of the user.

When the specific behavior is a tour behavior, because of a large number of countries, the user may travel a plurality of destinations (which may be countries), an LR model is established for each tour destination, and when the destination is estimated, a search feature is mainly used. Taking the example of training an LR model for predicting a user's outbound destination, the following outbound-related features of the LR algorithm and training samples are used: and (5) carrying out keyword training to obtain an LR model. The training process of the LR models corresponding to multiple destinations is the above process, and is not described herein again.

Two types of LR models are obtained through training, when the tour behavior of the user to be predicted needs to be predicted, if whether the user has the tour intention needs to be predicted, the LR model for predicting whether the user has the tour intention is used, the features meeting the input requirements of the LR model are extracted from the data of the user related to the tour, the extracted features are input into the LR model, and an output value is obtained, wherein the closer the output value is to 1, the more the user is likely to have the tour intention, the closer the output value is to 0, and the less the user is likely to have the tour intention.

If it is required to predict where the user goes to go for tour, a plurality of LR models for predicting the user's tour destination are used, each LR model corresponds to a tour destination, features satisfying the input requirements of the LR models are extracted from data of the user related to tour, the extracted features are input into the plurality of LR models to obtain a plurality of output values, wherein the closer the output value is to 1, the more likely the user goes to the destination corresponding to the output value, and the closer the output value is to 0, the less likely the user is to go to the destination corresponding to the output value.

As can be seen from the above embodiments, the embodiment may train a prediction model for predicting the specific behavior of the user according to the training samples and the specific algorithm, when the specific behavior of the user needs to be predicted, obtain data related to the specific behavior of the user, extract features related to the specific behavior from the data related to the specific behavior, input the features related to the specific behavior into the prediction model to obtain an output value, and predict the specific behavior of the user according to the output value. Because the data of the user related to the specific behavior can reflect the behavior intention of the user to a great extent, the embodiment of the specification can more accurately predict the specific behavior of the user, and then push the related information according to the predicted specific behavior, thereby improving the accuracy of pushing the information.

In addition, along with the development of internationalization, the increase of the users who travel abroad is selected, according to the behavior data of the users, the outbound intention of the users is intelligently obtained, marketing activities can be more effectively put in, physical examination of the users is promoted, the manual input time of operation can be reduced, the effect of operation activities is improved, the viscosity of products is increased, and the development of cross-country travel business is promoted. Based on this situation, in the embodiment of the present specification, the specific behavior may be a specific behavior for a country, and in this case, the data related to the specific behavior is data related to the country, and the feature related to the specific behavior is a feature related to the country. The LR model obtained by using the LR algorithm and the feature training related to the departure is used to predict whether the user has the intention of the departure and to predict which country the user wants to go, and the training process of the LR model is similar to that of the LR model in the embodiment shown in fig. 1, and therefore, the details are not repeated here.

The above describes the creation process of a prediction model for predicting a specific behavior of a user, and how to predict the specific behavior of the user using the created prediction model is described below.

FIG. 2 is a flow diagram of a model-based user behavior prediction method according to an embodiment of the present disclosure, which may include the following steps, as shown in FIG. 2:

in S202, data related to a specific behavior of the target user is acquired.

In this embodiment of the present specification, the target user is a user whose specific behavior is to be predicted, and the data of the target user related to the specific behavior is historical data of the target user related to the specific behavior.

In this embodiment, the data related to the specific behavior of the target user may include at least one of the following: search records, purchase records, collection records, browsing records, and application click times.

When the specific behavior is an outbound behavior, the data related to the outbound of the target user is historical data related to the outbound of the target user. The data of the target user related to the outing may include at least one of: search records, purchase records, collection records, browsing records, and application click times.

Optionally, in an embodiment, the S202 may include:

and acquiring data related to the specific behaviors of the target user within M days before the current time, wherein M is a natural number. And when the specific behavior is the tour behavior, acquiring data related to tour of the target user within M days before the current time.

In this embodiment, M may be a natural number set according to an empirical value, for example, M takes the value of 8; m may also be a value trained on neural networks.

In S204, features related to the specific behavior are extracted from the data related to the specific behavior of the target user.

In S206, the feature of the target user related to the specific behavior is input into a prediction model, which is obtained by training the feature of the training sample related to the specific behavior using a specific algorithm, to obtain an output value, and the prediction model is used to establish a mapping relationship between the feature related to the specific behavior and the prediction information of the specific behavior of the user.

Preferably, in the embodiment of the present specification, the specific algorithm is an LR algorithm, and the prediction model is an LR model. When using the LR model, it is necessary to extract features corresponding to the input request of the LR model from data on a specific behavior of a target user, and input the extracted features on the specific behavior that meet the input request of the target user into the LR model to obtain an output value. The output of the LR model is a probability value, and the value range of the probability value is 0-1.

In S208, the specific behavior of the target user is predicted from the output value.

As can be seen from the foregoing embodiments, when a specific behavior of a user needs to be predicted, the embodiment may acquire data related to the specific behavior of the user, extract features related to the specific behavior from the data related to the specific behavior, input the features related to the specific behavior into a prediction model to obtain an output value, and predict the specific behavior of the user according to the output value. Because the data of the user related to the specific behavior can reflect the behavior intention of the user to a great extent, the embodiment of the specification can more accurately predict the specific behavior of the user, and then push the related information according to the predicted specific behavior, thereby improving the accuracy of pushing the information.

In another embodiment provided by the present specification, whether a target user has an intention to make a specific behavior may be predicted according to data of the target user related to the specific behavior, in order to achieve the above object, a prediction model used is a model constructed according to at least two of search records, purchase records, collection records, browsing records, and application program click times of training samples, and the prediction model is used for predicting whether the user has the intention to make the specific behavior;

at this time, S202 in the embodiment shown in fig. 2 may specifically include:

and acquiring data which corresponds to the used prediction model and is relevant to the specific behaviors of the target user, wherein the type of the data which is relevant to the specific behaviors of the target user is the same as that of the data which is relevant to the specific behaviors of the training sample used for training the LR model.

At this time, S204 in the embodiment shown in fig. 2 may specifically include:

performing at least two of the following feature extraction operations: extracting keywords related to a specific behavior from a search record of a target user, extracting an article number and an article type related to the specific behavior from a purchase record of the target user, extracting an article number and an article type related to the specific behavior from a collection record of the target user, extracting an article number and an article type related to the specific behavior from a browsing record of the target user, and calculating a ratio of application program click times related to the specific behavior according to the application program click times of the target user;

and taking the content extracted by the characteristic extraction operation as the characteristic related to the specific behavior of the target user.

At this time, S206 in the embodiment shown in fig. 2 may specifically include:

and inputting the features which are extracted by the feature extraction operation, meet the input requirement of the prediction model and are related to the specific behaviors into the prediction model to obtain an output value.

At this time, S208 in the embodiment shown in fig. 2 may specifically include:

if the output value reaches a preset first threshold value, predicting that the target user has the intention of making a specific behavior;

and if the output value does not reach the preset first threshold value, predicting that the target user does not have the intention of making a specific behavior.

When the specific behavior is the tour behavior, the data related to the specific behavior is the data related to the tour, and the characteristics related to the specific behavior are the characteristics related to the tour; when the specific algorithm is an LR algorithm, the prediction model is an LR model; in this case, in one example, as shown in fig. 3, the LR model for predicting whether the user has the intention of going out is a model constructed according to the search record, the purchase record, the collection record, the browsing record and the number of clicks of the application, and the data related to going out of the target user should include: the method comprises the steps of searching records, purchasing records, collecting records, browsing records and application program clicking times of a target user, then extracting keywords related to the trip from the searching records of the target user, extracting an item number and an item type related to the trip from the purchasing records, collecting records and browsing records of the target user, calculating a proportion value of the application program clicking times related to the trip according to the application program clicking times of the target user, and inputting the extracted keywords, the item number, the item type and the proportion value into an LR model to obtain an output value, wherein the closer the output value is to 1, the more probable the target user has the trip intention, the closer the output value is to 0, the more probable the target user has the trip intention. The prediction result is specifically as follows: if the output value reaches a preset first threshold (for example, 0.7), predicting that the target user has the intention of going to a trip; and if the output value does not reach the preset first threshold value, predicting that the target user has no intention of going to a trip.

In another embodiment provided by the present specification, the intended specific behavior of the target user can be predicted according to the data related to the specific behavior of the target user, and in order to achieve the above object, the prediction model used includes: the method comprises the following steps that a first type of prediction model is built according to at least two of search records, purchase records, collection records, browsing records and application program clicking times of training samples, and a plurality of second type of prediction models are built according to the search records of the training samples, wherein the first type of prediction models are used for predicting whether a user has intention of making a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior;

at this time, S202 in the embodiment shown in fig. 2 may specifically include:

and acquiring data which corresponds to the used prediction model and is relevant to the specific behaviors of the target user, wherein the type of the data which is relevant to the specific behaviors of the target user is the same as that of the data which is relevant to the specific behaviors of the training sample used for training the prediction model.

At this time, S204 in the embodiment shown in fig. 2 may specifically include:

extracting keywords of the target user related to the specific behaviors from the search records of the target user;

if the first-class prediction model relates to the search record of the training sample in the construction process, performing at least one of the following feature extraction operations: extracting an article number and an article type related to a specific behavior from a purchase record of a target user, extracting an article number and an article type related to the specific behavior from a collection record of the target user, extracting an article number and an article type related to the specific behavior from a browsing record of the target user, and calculating a ratio of application program clicking times related to the specific behavior according to the application program clicking times of the target user;

if the search record of the training sample is not involved in the construction process of the first type of prediction model, at least two of the following feature extraction operations are executed: extracting an article number and an article type related to a specific behavior from a purchase record of a target user, extracting an article number and an article type related to the specific behavior from a collection record of the target user, extracting an article number and an article type related to the specific behavior from a browsing record of the target user, and calculating a ratio of application program clicking times related to the specific behavior according to the application program clicking times of the target user; and taking the keywords and the content extracted by the characteristic extraction operation as the characteristics related to the specific behaviors of the target user.

At this time, S206 in the embodiment shown in fig. 2 may specifically include:

if the construction process of the first-class prediction model relates to the search record of the training sample, inputting the keywords extracted from the search record of the target user and the content extracted by the characteristic extraction operation into the first-class prediction model to obtain an output value;

if the construction process of the first-class prediction model does not relate to the search record of the training sample, inputting the content extracted by the feature extraction operation into the first-class prediction model to obtain an output value;

and if the output value of the first-class prediction model is larger than a preset second threshold value, respectively inputting the keywords extracted from the search records of the target user into a plurality of second-class prediction models to obtain a plurality of output values.

At this time, S208 in the embodiment shown in fig. 2 may specifically include:

determining output values of N bits before the value ranking from a plurality of output values corresponding to the second type of prediction model;

and determining the specific behavior corresponding to the output value of N bits before the value ranking as the intention specific behavior of the target user, wherein N is a natural number. In practical applications, N may take the value of 5.

When the specific behavior is the tour behavior, the data related to the specific behavior is the data related to the tour, and the characteristics related to the specific behavior are the characteristics related to the tour; when the specific algorithm is an LR algorithm, the prediction model is an LR model; at this time, in an example, as shown in fig. 4, a process of inputting characteristics related to the trip of the target user into an LR model for predicting whether the user has the trip intention is similar to the previous embodiment, and details thereof are omitted. In fig. 4, 4 second-class LR models are shown, respectively: an LR model a for predicting the user's outgoing destination, an LR model B for predicting the user's outgoing destination, an LR model C for predicting the user's outgoing destination, and an LR model D for predicting the user's outgoing destination, and a keyword extracted from a search record of a target user is input to the 4 second-type LR models to obtain 4 output values, for example, 4 output values are: 0.9, 0.95, 0.6 and 0.5, and predicting the destinations corresponding to the output values of the first two values, namely the destinations corresponding to the output values of 0.9 and 0.95, as the outbound destinations of the target user.

Along with the development of the internationalization, the number of the users who visit the country is increased, according to the behavior data of the users, the visiting intention of the users is intelligently obtained, marketing activities can be more effectively put in, physical examination of the users is promoted, the manual input time of the operation can be reduced, the effect of the operation activities is improved, the viscosity of products is further increased, and the development of cross-country tour business is promoted.

Based on this situation, in an embodiment of the present specification, the tour may be a outbound tour, in this case, in another embodiment provided by the present specification, the LR model used is an LR for predicting the outbound tour behavior of the target user, and the data related to the tour of the training samples include: the data related to the foreign country of the training sample, the characteristics related to the outing of the training sample comprise: the feature of the training sample related to the country is trained by using the LR algorithm and the feature of the training sample related to the country to obtain an LR model, which can be used to predict whether the user has a intention to go to the country and to predict which country the user wants to go to.

When the LR model is used to predict the outbound behavior of the target user, it is necessary to obtain data of the target user related to the outbound, extract features of the target user related to the outbound, input the LR model to obtain an output value, and predict the outbound behavior of the target user according to the output value, which is not described herein again because the prediction process is similar to that of the embodiment shown in fig. 2.

In the embodiment of the specification, the prediction result can be used for an intelligent marketing scheme before international travel so as to ensure that high-value potential outbound population is excavated, and further, different operation activities can be configured for different countries, so that the efficiency and effect of the operation activities are effectively improved.

Therefore, the method and the device can predict potential foreign people based on the LR model, so that high precision is achieved, the LR algorithm is simple, the calculation complexity is low, and the method and the device can deal with ultra-large-scale user behavior data. In addition, after the potential foreign people results are based, the travel destination of the potential foreign people is further predicted, the intention of the people is further refined, and the accuracy and the effect of operation activities are effectively improved.

Fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present specification. Referring to fig. 5, at a hardware level, the electronic device includes a processor, and optionally further includes an internal bus, a network interface, and a memory. The Memory may include a Memory, such as a Random-Access Memory (RAM), and may further include a non-volatile Memory, such as at least 1 disk Memory. Of course, the electronic device may also include hardware required for other services.

The processor, the network interface, and the memory may be connected to each other via an internal bus, which may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one double-headed arrow is shown in FIG. 5, but this does not indicate only one bus or one type of bus.

And the memory is used for storing programs. In particular, the program may include program code comprising computer operating instructions. The memory may include both memory and non-volatile storage and provides instructions and data to the processor.

The processor reads the corresponding computer program from the nonvolatile memory into the memory and then runs the computer program to form the model training device on the logic level. The processor is used for executing the program stored in the memory and is specifically used for executing the following operations:

The method performed by the model training apparatus disclosed in the embodiment of fig. 5 in this specification can be applied to or implemented by a processor. The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present specification may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present specification may be embodied directly in a hardware decoding processor, or in a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor.

The electronic device may also execute the method of fig. 1 and implement the functions of the model training apparatus in the embodiment shown in fig. 1, which are not described herein again in this specification.

Fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present specification. Referring to fig. 6, at a hardware level, the electronic device includes a processor, and optionally further includes an internal bus, a network interface, and a memory. The Memory may include a Memory, such as a Random-Access Memory (RAM), and may further include a non-volatile Memory, such as at least 1 disk Memory. Of course, the electronic device may also include hardware required for other services.

The processor, the network interface, and the memory may be connected to each other via an internal bus, which may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one double-headed arrow is shown in FIG. 6, but that does not indicate only one bus or one type of bus.

The processor reads a corresponding computer program from the nonvolatile memory into the memory and then runs the computer program to form a model-based user behavior prediction device on a logic level. The processor is used for executing the program stored in the memory and is specifically used for executing the following operations:

acquiring data related to specific behaviors of a target user;

The method performed by the model-based user behavior prediction apparatus according to the embodiment shown in fig. 6 of the present specification may be applied to or implemented by a processor. The processor may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware in a processor or instructions in the form of software. The Processor may be a general-purpose Processor, including a Central Processing Unit (CPU), a Network Processor (NP), and the like; but also Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present specification may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present specification may be embodied directly in a hardware decoding processor, or in a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory, and a processor reads information in the memory and completes the steps of the method in combination with hardware of the processor.

The electronic device may also execute the method of fig. 2, and implement the functions of the model-based user behavior prediction apparatus in the embodiment shown in fig. 2, which are not described herein again in this specification.

Of course, besides the software implementation, the electronic device in this 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 also be hardware or logic devices.

The present specification embodiments also provide a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a portable electronic device comprising a plurality of application programs, enable the portable electronic device to perform the method of the embodiment shown in fig. 1, and in particular to perform the method of:

The present specification embodiments also provide a computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a portable electronic device comprising a plurality of application programs, enable the portable electronic device to perform the method of the embodiment shown in fig. 2, and in particular to perform the method of:

acquiring data related to specific behaviors of a target user;

Fig. 7 is a schematic structural diagram of a model training apparatus according to an embodiment of the present disclosure, and referring to fig. 7, in a software implementation, the model training apparatus 700 may include:

a first obtaining unit 701, configured to obtain a training sample set, where the training sample set includes training samples for training a model, and the training samples include data of a user related to a specific behavior;

a first extracting unit 702, configured to extract features of training samples in the training sample set acquired by the first acquiring unit 701, where the features are related to a specific behavior;

a training unit 703, configured to train the features related to the specific behavior extracted by the first extraction unit 702 according to a specific algorithm, so as to obtain a prediction model, where the prediction model is used to establish a mapping relationship between the features related to the specific behavior and the specific behavior prediction information of the user.

Optionally, as an embodiment, the data related to the specific behavior includes at least one of: search records, purchase records, collection records, browsing records, and application click times.

Optionally, as an embodiment, the data related to the specific behavior includes: searching for records;

the first extraction unit 702 includes:

a first extraction subunit, configured to extract a keyword related to a specific behavior from the search record;

a first feature determination subunit for regarding the keyword as one of a feature related to a specific behavior or a feature related to a specific behavior.

Optionally, as an embodiment, the data related to the specific behavior includes: recording the purchase;

the first extraction unit 702 includes:

the second extraction subunit is used for extracting the item number and the item type related to the specific behavior from the purchase record;

and the second characteristic determining subunit is used for taking the item number and the item type as one of the characteristics related to the specific behavior or the characteristics related to the specific behavior.

Optionally, as an embodiment, the data related to the specific behavior includes: collecting records;

the first extraction unit 702 includes:

the third extraction subunit is used for extracting the item number and the item type related to the specific behavior from the collection record;

and the third characteristic determining subunit is used for taking the item number and the item type as one of the characteristics related to the specific behavior or the characteristics related to the specific behavior.

Optionally, as an embodiment, the data related to the specific behavior includes: browsing the records;

the first extraction unit 702 includes:

the fourth extraction subunit is used for extracting the item number and the item type related to the specific behavior from the browsing record;

and the fourth characteristic determining subunit is used for taking the item number and the item type as one of the characteristics related to the specific behavior or the characteristics related to the specific behavior.

Optionally, as an embodiment, the data related to the specific behavior includes: the number of clicks of the application;

the first extraction unit 702 includes:

the fifth extraction subunit is used for calculating the ratio of the application program clicking times related to the specific behavior according to the application program clicking times;

a fifth feature determination subunit operable to take the ratio value as one of a feature related to the specific behavior or a feature related to the specific behavior.

Optionally, as an embodiment, the data related to the specific behavior further includes: the system comprises user basic information and a user tag, wherein the user tag is used for representing whether the user is a specific behavior fan;

the first extraction unit 702 further includes:

a sixth extraction subunit, configured to extract a user basic information feature from the user basic information;

and the sixth characteristic determining subunit is used for taking the user basic information characteristic and the user label as one of the characteristics related to the specific behavior.

Optionally, as an embodiment, the specific algorithm is a logistic regression LR algorithm, and the prediction model is an LR model.

Optionally, as an embodiment, the specific behavior is an outbound behavior, the data related to the specific behavior is data related to outbound, and the feature related to the specific behavior is a feature related to outbound.

The model training apparatus 700 may also execute the method of the embodiment shown in fig. 1, and implement the functions of the model training apparatus in the embodiment shown in fig. 7, which are not described herein again in this specification.

Fig. 8 is a schematic structural diagram of a model-based user behavior prediction device according to an embodiment of the present disclosure, and referring to fig. 8, in a software implementation, the model-based user behavior prediction device 800 may include:

a second acquisition unit 801 configured to acquire data of a target user relating to a specific behavior;

a second extracting unit 802, configured to extract features related to a specific behavior from the data related to the specific behavior acquired by the second acquiring unit 801;

a processing unit 803, configured to input the features related to the specific behavior extracted by the second extraction unit 802 into a prediction model, which is obtained by training the features related to the specific behavior of the training sample using a specific algorithm, to obtain an output value, where the prediction model is used to establish a mapping relationship between the features related to the specific behavior and prediction information of the specific behavior of the user;

a predicting unit 804, configured to predict a specific behavior of the target user according to the output value processed by the processing unit 803.

Optionally, as an embodiment, the prediction model is a model constructed according to at least two of search records, purchase records, collection records, browsing records and application click times, and is used for predicting whether the user has an intention to make a specific behavior;

the second extraction unit 802 includes:

a seventh extraction subunit for performing at least two of the following feature extraction operations: extracting keywords related to a specific behavior from the search records, extracting an article number and an article type related to the specific behavior from the purchase records, extracting an article number and an article type related to the specific behavior from the collection records, extracting an article number and an article type related to the specific behavior from the browsing records, and calculating a ratio of application program clicking times related to the specific behavior according to the application program clicking times;

and a seventh feature determining subunit, configured to use the content extracted by the feature extraction operation as a feature related to a specific behavior.

Optionally, as an embodiment, the prediction unit 804 includes:

the first predictor is used for predicting the intention of the target user to make a specific action under the condition that the output value reaches a preset first threshold value; and under the condition that the output value does not reach the preset first threshold value, predicting that the target user does not have the intention of making a specific behavior.

Optionally, as an embodiment, the prediction model includes: the system comprises a first type of prediction model and a plurality of second type of prediction models, wherein the first type of prediction models are constructed according to at least two of search records, purchase records, collection records, browsing records and application program clicking times, the second type of prediction models are constructed according to the search records, the first type of prediction models are used for predicting whether a user has intention to make a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior;

the second extraction unit 802 includes:

an eighth extraction subunit, configured to extract a keyword related to a specific behavior from the search record; and the number of the first and second groups,

if the first-class prediction model relates to a search record in the construction process, performing at least one of the following feature extraction operations: extracting the article number and the article type related to the specific behavior from the purchase record, extracting the article number and the article type related to the specific behavior from the collection record, extracting the article number and the article type related to the specific behavior from the browsing record, and calculating the ratio of the application program clicking times related to the specific behavior according to the application program clicking times;

if the first type of prediction model does not relate to the search record in the construction process, at least two of the following feature extraction operations are executed: extracting the article number and the article type related to the specific behavior from the purchase record, extracting the article number and the article type related to the specific behavior from the collection record, extracting the article number and the article type related to the specific behavior from the browsing record, and calculating the ratio of the application program clicking times related to the specific behavior according to the application program clicking times;

an eighth feature determining subunit, configured to use the keyword and the content extracted by the feature extraction operation as features related to a specific behavior;

the processing unit 803 includes:

the first processing subunit is used for inputting the keywords and the content extracted by the feature extraction operation into the first-class prediction model to obtain an output value under the condition that the construction process of the first-class prediction model relates to search records; and the number of the first and second groups,

under the condition that the construction process of the first-class prediction model does not involve searching records, inputting the content extracted by the feature extraction operation into the first-class prediction model to obtain an output value;

and the second processing subunit is used for respectively inputting the keywords into the plurality of second-class prediction models to obtain a plurality of output values under the condition that the output values are greater than a preset second threshold value.

Optionally, as an embodiment, the prediction unit 804 includes:

the output value screening submenu is used for determining an output value of N bits before the value ranking from a plurality of output values corresponding to the second type of prediction model;

and the second predictor unit is used for determining the specific behavior corresponding to the output value of N bits before the value ranking as the intention specific behavior of the target user, wherein N is a natural number.

Optionally, as an embodiment, the second obtaining unit 801 includes:

and acquiring data related to specific behaviors of the target user within M days before the current time, wherein M is a natural number.

Optionally, as an embodiment, the specific algorithm is an LR algorithm, and the predictive model is an LR model.

The model-based user behavior prediction apparatus 800 may further perform the method in the embodiment shown in fig. 2, and implement the functions of the model-based user behavior prediction apparatus in the embodiment shown in fig. 8, which are not described herein again in this specification.

In short, the above description is only a preferred embodiment of the present disclosure, and is not intended to limit the scope of the present disclosure. Any modification, equivalent replacement, improvement and the like made within the spirit and principle of the present specification shall be included in the protection scope of the present specification.

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.

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.

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.

Claims

1. A method of model training, the method comprising:

obtaining a training sample set, wherein the training sample set comprises training samples for training a model, the training samples comprise data related to specific behaviors of a user, and the data related to the specific behaviors comprises at least one of the following data: searching records, purchasing records, collecting records, browsing records and application program clicking times;

training the features related to the specific behaviors according to a specific algorithm to obtain a prediction model, wherein the prediction model is used for establishing a mapping relation between the features related to the specific behaviors and the specific behavior prediction information of the user, and the prediction model comprises the following steps: the system comprises a first type of prediction model and a plurality of second type of prediction models, wherein the first type of prediction models are constructed according to at least two of search records, purchase records, collection records, browsing records and application program click times, the plurality of second type of prediction models are constructed according to the search records, the first type of prediction models are used for predicting whether a user has intention to make a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior.

2. The method of claim 1, the data related to a particular behavior comprising at least one of: search records, purchase records, collection records, browsing records, and application click times.

3. The method of claim 2, the data related to a particular behavior comprising: searching for records;

the extracting of the features related to the specific behaviors of the training samples in the training sample set comprises:

extracting keywords related to a specific behavior from the search records;

the keyword is taken as one of a feature related to a specific behavior or a feature related to a specific behavior.

4. The method of claim 2, the data related to a particular behavior comprising: recording the purchase;

extracting an item number and an item category related to a specific action from the purchase record;

and taking the item number and the item category as one of the characteristics related to the specific action or the characteristics related to the specific action.

5. The method of claim 2, the data related to a particular behavior comprising: collecting records;

extracting the item numbers and the item types related to specific behaviors from the collection records;

6. The method of claim 2, the data related to a particular behavior comprising: browsing the records;

7. The method of claim 2, the data related to a particular behavior comprising: the number of clicks of the application;

the percentage value is taken as one of the characteristic related to the specific behavior or the characteristic related to the specific behavior.

8. The method of any of claims 2-7, the data related to a particular behavior further comprising: the system comprises user basic information and a user tag, wherein the user tag is used for representing whether the user is a specific behavior fan;

the extracting features of the training samples in the training sample set related to the specific behaviors further comprises:

and extracting the basic information features of the user from the basic information of the user, and taking the basic information features of the user and the user label as one of features related to specific behaviors.

9. The method of claim 1, wherein the specific algorithm is a Logistic Regression (LR) algorithm and the predictive model is an LR model.

10. The method of claim 1, wherein the specific behavior is an outbound behavior, the data related to the specific behavior is data related to an outbound, and the feature related to the specific behavior is a feature related to an outbound.

11. A model-based user behavior prediction method, the method comprising:

acquiring data related to specific behaviors of a target user;

extracting features related to a specific behavior from the data related to the specific behavior, the data related to the specific behavior including at least one of: searching records, purchasing records, collecting records, browsing records and application program clicking times;

inputting the features related to the specific behaviors into a prediction model to obtain an output value, wherein the prediction model is obtained by training the features related to the specific behaviors of a training sample by using a specific algorithm, and is used for establishing a mapping relation between the features related to the specific behaviors and the specific behavior prediction information of a user, and the prediction model comprises: the system comprises a first type of prediction model and a plurality of second type of prediction models, wherein the first type of prediction models are constructed according to at least two of search records, purchase records, collection records, browsing records and application program clicking times, the second type of prediction models are constructed according to the search records, the first type of prediction models are used for predicting whether a user has intention to make a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior;

12. The method of claim 11, wherein the prediction model is a model constructed according to at least two of search records, purchase records, collection records, browsing records, and application click times, and is used for predicting whether a user has an intention to make a specific behavior;

the extracting of the features related to the specific behaviors from the data related to the specific behaviors comprises:

performing at least two of the following feature extraction operations: extracting keywords related to a specific behavior from the search records, extracting an article number and an article type related to the specific behavior from the purchase records, extracting an article number and an article type related to the specific behavior from the collection records, extracting an article number and an article type related to the specific behavior from the browsing records, and calculating a ratio of application program clicking times related to the specific behavior according to the application program clicking times;

and taking the content extracted by the characteristic extraction operation as the characteristic related to the specific behavior.

13. The method of claim 12, said predicting, from the output value, a particular behavior of the target user, comprising:

14. The method of claim 11, wherein the step of selecting the target,

extracting keywords related to a specific behavior from the search records;

taking the keywords and the content extracted by the feature extraction operation as features related to specific behaviors;

inputting the features related to the specific behaviors into a prediction model to obtain output values, wherein the output values comprise:

if the construction process of the first type of prediction model relates to search records, inputting the keywords and the content extracted by the feature extraction operation into the first type of prediction model to obtain an output value;

if the construction process of the first-class prediction model does not involve search records, inputting the content extracted by the feature extraction operation into the first-class prediction model to obtain an output value;

and if the output value is larger than a preset second threshold value, respectively inputting the keywords into the plurality of second-class prediction models to obtain a plurality of output values.

15. The method of claim 14, said predicting, from the output value, a particular behavior of the target user, comprising:

and determining the specific behavior corresponding to the output value of N bits before the value ranking as the intention specific behavior of the target user, wherein N is a natural number.

16. The method of claim 11, the obtaining data related to a particular behavior of a target user, comprising:

17. The method of claim 11, wherein the specific algorithm is an LR algorithm and the predictive model is an LR model.

18. The method of claim 11, wherein the specific behavior is an outbound behavior, the data related to the specific behavior is data related to an outbound, and the feature related to the specific behavior is a feature related to an outbound.

19. A model training apparatus, the apparatus comprising:

a first obtaining unit, configured to obtain a training sample set, where the training sample set includes training samples for training a model, where the training samples include data related to a specific behavior of a user, and the data related to the specific behavior includes at least one of: searching records, purchasing records, collecting records, browsing records and application program clicking times;

a training unit, configured to train, according to a specific algorithm, the features related to the specific behavior extracted by the first extraction unit, so as to obtain a prediction model, where the prediction model is used to establish a mapping relationship between the features related to the specific behavior and prediction information of the specific behavior of the user, and the prediction model includes: the system comprises a first type of prediction model and a plurality of second type of prediction models, wherein the first type of prediction models are constructed according to at least two of search records, purchase records, collection records, browsing records and application program click times, the plurality of second type of prediction models are constructed according to the search records, the first type of prediction models are used for predicting whether a user has intention to make a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior.

20. An apparatus for model-based user behavior prediction, the apparatus comprising:

a second obtaining unit, configured to obtain data related to a specific behavior of a target user, where the data related to the specific behavior includes at least one of: searching records, purchasing records, collecting records, browsing records and application program clicking times;

a processing unit, configured to input the features related to the specific behavior extracted by the second extraction unit into a prediction model to obtain an output value, where the prediction model is a model obtained by training the features related to the specific behavior of the training sample using a specific algorithm, and the prediction model is used to establish a mapping relationship between the features related to the specific behavior and prediction information of the specific behavior of the user, and the prediction model includes: the system comprises a first type of prediction model and a plurality of second type of prediction models, wherein the first type of prediction models are constructed according to at least two of search records, purchase records, collection records, browsing records and application program clicking times, the second type of prediction models are constructed according to the search records, the first type of prediction models are used for predicting whether a user has intention to make a specific behavior, the second type of prediction models are used for predicting the intention specific behavior of the user, and one second type of prediction model corresponds to one intention specific behavior;

21. An electronic device, comprising:

a processor; and

22. An electronic device, comprising:

a processor; and

acquiring data related to specific behaviors of a target user, wherein the data related to the specific behaviors comprises at least one of the following data: searching records, purchasing records, collecting records, browsing records and application program clicking times;

23. A computer storage medium storing one or more programs that, when executed by an electronic device including a plurality of application programs, cause the electronic device to:

24. A computer storage medium storing one or more programs that, when executed by an electronic device including a plurality of application programs, cause the electronic device to: