CN110163703B

CN110163703B - Classification model establishing method, file pushing method and server

Info

Publication number: CN110163703B
Application number: CN201810145655.6A
Authority: CN
Inventors: 周济民; 黄恒; 严玉良; 郎君; 司罗
Original assignee: Alibaba Group Holding Ltd
Current assignee: Alibaba Group Holding Ltd
Priority date: 2018-02-12
Filing date: 2018-02-12
Publication date: 2022-07-29
Anticipated expiration: 2038-02-12
Also published as: CN110163703A

Abstract

The application provides a classification model establishing method, a document pushing method and a server, wherein the classification model establishing method comprises the following steps: acquiring a plurality of purchasing behavior data and a plurality of reading behavior data; extracting users with purchasing behaviors and reading behaviors from the purchasing behavior data and the reading behavior data as training samples; and training to obtain a classification model through the positive sample, wherein the classification model is used for determining whether a reading file needs to be pushed to a target user according to the behavior data of the user. By means of the mode, the technical problems that an existing file pushing effect is not good and accuracy is low are solved, and the technical effect of accurately and efficiently pushing files is achieved.

Description

Classification model establishing method, file pushing method and server

Technical Field

The application belongs to the technical field of internet, and particularly relates to a classification model establishing method, a document pushing method and a server.

Background

More and more people are currently purchasing the desired items through e-commerce shopping platforms. However, when a user makes a shopping decision on an e-commerce shopping platform, the user is often limited by insufficient knowledge and experience, and needs to go to other websites (e.g., a search website or a science popularization website) to inquire about data related to an item desired to be purchased so as to help the user make a decision.

Consider that if the e-commerce shopping platform is transferred to other websites for data query, the shopping link is often lengthened, resulting in a poor user experience. In order to make the user get a more satisfactory shopping experience, the e-commerce website often provides knowledge of the relevant fields of the items that the user wishes to buy or the shopping thoughts of other users, and the knowledge and the shopping thoughts can help the user to make a shopping decision.

However, the e-commerce shopping platform is used as a platform for selling goods, and articles are not displayed by too many resources. Aiming at the problems of accurately and effectively pushing articles to a user and not influencing the article sale of an e-commerce shopping platform, an effective solution is not provided at present.

Disclosure of Invention

The application aims to provide a classification model establishing method, a document pushing method and a server, and the purpose of simply and accurately pushing a document to a user to help the user make a shopping decision can be achieved.

The application provides a classification model establishing method, a document pushing method and a server, which are realized as follows:

a classification model building method, the method comprising:

acquiring a plurality of purchasing behavior data and a plurality of reading behavior data;

Extracting users with purchasing behaviors and reading behaviors from the purchasing behavior data and the reading behavior data to serve as training samples;

and training to obtain a classification model through the training sample, wherein the classification model is used for determining whether the reading file needs to be pushed to a target user according to the behavior data of the user.

A document push method, the method comprising:

acquiring behavior data and user characteristics of a user;

determining whether a file needs to be pushed to the user or not through a preset classification model;

and under the condition that the file needs to be pushed, pushing the file to the user.

A server comprising a processor and a memory for storing processor-executable instructions, the instructions when executed by the processor implementing the steps of the above method.

A computer readable storage medium having stored thereon computer instructions which, when executed, implement the steps of the above-described method.

According to the classification model establishing method, the file pushing method and the server, the association degree between the purchasing behavior and the reading behavior of the user is established, so that the group which purchases and is willing to read the file can be effectively determined, and the pushing value and the clicked rate of the pushed file can be improved. By means of the mode, the technical problems that an existing file pushing effect is not good and accuracy is low are solved, and the technical effect of accurately and efficiently pushing files is achieved.

Drawings

In order to more clearly illustrate the embodiments of the present application 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 application, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort.

FIG. 1 is a schematic diagram of a document pushing system according to an embodiment of the present application;

FIG. 2 is a flowchart of a method of a document pushing method according to an embodiment of the present application;

FIG. 3 is a schematic diagram of a document push interface according to an embodiment of the present application;

FIG. 4 is another schematic diagram of a document push interface according to an embodiment of the present application;

FIG. 5 is a flowchart of a document push algorithm according to an embodiment of the present application;

FIG. 6 is a schematic diagram of a document pushing architecture according to an embodiment of the present application;

fig. 7 is a schematic diagram of a model structure of a document pushing system according to an embodiment of the present application.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

It is considered that a suitable article can be pushed to a user as a recommendation problem, namely, a problem of personalized article recommendation to the user in a suitable scene. However, for the e-commerce shopping platform, the article reading behavior of the user is very little, even for most users, the article reading behavior is absent, for a certain user, it is difficult to obtain the preference data of the user, and it is difficult to achieve an ideal effect by adopting a general processing mode of recommendation problems.

Based on this, it is considered that if the reading behaviors of some users with reading behaviors can be found from massive user behaviors, and the relationship between the reading behaviors and the commodity searched and purchased by the user is also considered. The data may be combined to recommend documentation for other users on the platform. In the case of few reading behaviors of the user, the method in the example is better in accuracy and adaptability compared with the existing method of recommending articles based on the reading behavior of the target user only. Compared with the prior art that the article is recommended only according to the browsing behavior of the target user on the commodity, the method in the embodiment can reduce the pushing of unnecessary articles, so that the article pushing accuracy is higher.

To this end, in this example, a document pushing system is provided, as shown in fig. 1, which may include: the system comprises a client and a server, wherein the server can push a file to the client.

In one embodiment, the client may be a terminal device or software used by a client operation. Specifically, the client may be a terminal device such as a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart watch, or other wearable devices. Of course, the client may also be software that can run in the terminal device. For example: shopping, sea panning and other application software. The server can push the file to the client to realize the pushing of the file, and the user receives and checks the file through the client.

In an embodiment, the server may be a single server device, a server cluster, a cloud server, or the like. The specific manner adopted can be selected according to actual needs, and the application does not limit the manner.

Based on the above-mentioned document pushing system, as shown in fig. 2, a document pushing method is provided to recommend a document to a user in an e-commerce environment, so as to help the user make a shopping decision. For example, the push may be performed while the user is browsing merchandise, so as to help the user make decisions. Or pushing when the user browses the goods but does not make a purchase, so as to give the user a purchase reference.

It should be noted that, in this example, the document may be an article, a set of articles, a title of an article or a set of articles, a link of an article, a link of a set of articles, or the like, which is not limited in this application.

Although the present application provides method operational steps or apparatus configurations as illustrated in the following examples or figures, more or fewer operational steps or modular units may be included in the methods or apparatus based on conventional or non-inventive efforts. In the case of steps or structures which do not logically have the necessary cause and effect relationship, the execution sequence of the steps or the module structure of the apparatus is not limited to the execution sequence or the module structure described in the embodiments and shown in the drawings of the present application. When the described method or module structure is applied in an actual device or end product, the method or module structure according to the embodiments or shown in the drawings can be executed sequentially or executed in parallel (for example, in a parallel processor or multi-thread processing environment, or even in a distributed processing environment).

As shown in fig. 2, the method for pushing the document of this embodiment may include the following steps:

Step 201: acquiring a plurality of purchasing behavior data and a plurality of reading behavior data;

specifically, when implemented, some user behavior data other than purchasing behavior data may also be obtained, which may include, but is not limited to, one or more of the following: the historical search behavior data, the historical purchase behavior data, the historical collection behavior data, the historical attention behavior data, the historical purchase wish single-line data and the like are used as reference data for training. In actual implementation, the type of the corresponding historical behavior data may be selected according to actual needs, which is not limited in the present application.

In one embodiment, the purchasing behavior data and the reading behavior data of a large number of users may be analyzed, for example, historical purchasing behavior data, historical search behavior data, and reading behavior data corresponding to the respective historical search behavior data and historical purchasing behavior data of a plurality of users may be obtained. That is, after searching for a certain product, the user reads the article and then purchases the product, or after searching for a certain product, the user directly purchases the product without reading the article. And taking the acquired data as a basis for subsequently pushing the file to the target user.

In one embodiment, when the historical reading behavior data and the corresponding reading behavior data are obtained, the data of the article reading faithful user can be obtained. When the method is implemented, the users meeting certain conditions can be used as article reading faithful users. For example, a user who reads an article for multiple times and has an article reading time exceeding a preset time length may be used as an article reading faithful user, and specifically, for example, a user who reads an article for more than or equal to three times in a month may be used as an article reading faithful user, and a user whose reading time is above 1 minute each time may be used as an article reading faithful user.

However, it should be noted that the above-listed determination conditions for the article reading loyalty users are only an exemplary description, and other manners or data thresholds may be used for determination when the determination is actually implemented, and the present application is not limited thereto.

Of course, in implementation, a large amount of data may be obtained by clustering from a large amount of user data, or users meeting the requirement of reading faithful users of the article may be screened or mined in advance, and then the data of the users may be obtained. The specific method can be selected according to actual needs, and the application does not limit the method.

For example, user data may be obtained for a predetermined period of time (e.g., data for the last 30 days), wherein the user data may include: the behavior data of commodity searching, collecting, purchasing and the like and the behavior data of whether articles are read or not. When data is acquired, a distributed computing engine can be adopted to acquire the data in consideration of the large data volume.

Step 202: extracting users with purchasing behaviors and reading behaviors from the purchasing behavior data and the reading behavior data as training samples; training to obtain a classification model through the training sample, wherein the classification model is used for determining whether reading documentations need to be pushed to a target user according to behavior data of the user;

in one embodiment, the relationship between the search purchasing behavior and the reading behavior is established based on the acquired data, the relationship between the search purchasing behavior and the reading behavior is established in a cluster analysis manner, or a training model is established based on the acquired data to train the model, so as to obtain a training model for representing the relationship between the search purchasing behavior and the reading behavior.

In order to establish the relationship between the search purchasing behavior and the reading behavior, it needs to be determined which search purchasing behavior is reading behavior and which search purchasing behavior is not reading behavior. In one embodiment, in order to determine which search purchasing behavior is reading behavior, the historical behavior of each training user may be sorted according to time, when the behavior of reading articles is encountered, a predetermined time (for example, 5 days) is advanced from the current time, the categories of products searched or browsed and purchased in five days are discretized into the predetermined time (for example, three intervals of 1 day, 1 to 3 days and more than 3 days), and then whether the search purchasing behavior corresponding to the read articles exists or not is determined based on the search purchasing behavior. Where correspondence may be finding the same or similar categories.

For example: the categories for the articles read are: the mobile phone shell, the mobile phone film and the like can be used as categories corresponding to the read article. When the method is implemented, the category corresponding to the search purchasing behavior data and the article reading behavior data serving as the sample is ensured to be consistent.

In one embodiment, the reading behavior data may include, but is not limited to, at least one of: whether to read the article and the product category corresponding to the read article.

In one embodiment, the historical behavior data may include: user characteristics of the user, categories of search products, categories of purchase products, and categories related to categories of purchase products or categories of search products. Wherein the user characteristics may include, but are not limited to, at least one of: gender, age, location, purchasing power, category preferences.

After the historical behavior data and the reading behavior data are obtained, each piece of data can be formed into a sample to facilitate clustering or modeling. For example, one sample may be obtained as follows:

1) and the behavior characteristics of searching, collecting and purchasing the commodities within 5 days.

2) Characteristics such as user age, gender, region, purchasing power, family, married, fertile, category preference, etc.;

3) Similar categories (which may be derived from the similarity between categories).

4) Whether there is reading behavior.

For sample data, not only data with reading behaviors but also data without reading behaviors are required, and the data are combined for judgment, so that whether the file needs to be pushed to a target object can be determined based on the height of a threshold value during judgment.

Step 203: and determining whether a reading file needs to be pushed to the target user or not through the classification model.

In one embodiment, determining whether the reading file needs to be pushed to the target user through the classification model may include:

s1: acquiring search behavior data of a target user;

s2: acquiring user characteristics of the target user;

s3: determining whether to push a file to a user and the reading interest category of the target user according to the established relation between the historical behavior data and the reading behavior, and the behavior data and the user characteristics of the target user;

s4: and pushing the file related to the reading interest category to the target user.

That is, real-time behavior data of the target user, or behavior data of the target user in a current period of time, and user characteristics (e.g., gender, age, region, purchasing ability, category preference) of the target user may be obtained. This is similar to the previous acquisition of historical behavior data of multiple users, and after the data is acquired, the user may be determined to be recommended or an article category that the user wishes to be recommended according to the established relationship between purchasing behavior and reading behavior. For example, if it is determined that the category that the user wishes or needs to be recommended is "dress," a case for "dress" may be recommended for the user. For example: "how to choose the one-piece dress, the old driver with you fly", after clicking the case, the user can display a plurality of articles related to the case.

In order to avoid that the user experience is low when the file is pushed to the user at an improper time, or that the file is pushed to the user after the user purchases a certain product, the pushed file cannot play a due role. In one embodiment, the document may be pushed to the target user in the case that it is determined that the user has a search behavior within a predetermined period of time but has no purchase behavior matching the search behavior.

When the file is pushed, the file can be pushed on a search browsing page of a target user, or the file can be pushed to the target user in an application information mode.

As shown in fig. 3, an interface schematic diagram for pushing a document on a search browsing page of a target user; as shown in fig. 4, the schematic interface diagram is an interface diagram for pushing a document to a target user in a manner of applying information.

However, it should be noted that the above listed push method is only a schematic description, and in practical implementation, other push methods may be adopted, for example, text push may be performed by a short message or a small icon, and specifically, which method is adopted may be selected according to actual needs, which is not limited in this application.

The above document pushing method is described below with reference to a specific scenario, however, it should be noted that this specific embodiment is only for better describing the present application, and does not constitute a specific limitation to the present application.

In the implementation scenario, model training is used as a file pushing mode, and other modes such as cluster analysis and the like can be adopted in implementation, which is not limited in the present application.

As shown in fig. 5, the following steps may be included:

s1: reading faithful users of the mined article:

in this example, the user who reads the article multiple times (three times or more) in one month and whose reading stay time exceeds a certain threshold value is regarded as an article reading faithful user.

S2: mining historical behavior data and reading behavior data of article reading faithful users:

for example, behavior data of product search, collection, shopping cart adding, purchase and the like of each user in the article reading faithful users within the last 30 days and behavior data of the articles read by the users can be acquired.

After the data are obtained, the behavior data of each user can be sorted according to time, the data are searched from front to back, if the reading behavior data are met, the data can be pushed forward for 5 days from the current time, the commodity category behaviors in 5 days are discretized into three intervals of 1 day, 1 to 3 days and more than 3 days, and the times are counted according to behavior types. In the statistical time, the categories of the product searching behavior and the content reading behavior need to be ensured to be consistent: the similarity of the query of the searched commodity and the query of the searched content is calculated to be guaranteed within a certain threshold value, so that the category consistency of the product searching behavior and the content reading behavior is guaranteed.

The reading behavior data obtained in each time is used as a training sample, and the training sample may include the following characteristics:

4) Whether there is reading behavior.

S3: training a model:

the process of training the model in this example can be understood as a process of establishing a degree of correlation between the purchasing behavior and the reading behavior of the user, i.e. establishing a classifier by which it can be identified which users are those who need to push the document and which users are those who do not need to push the document.

For this reason, in order to train the classifier, the forward training samples required for training are: a user who purchases an item and has a reading behavior. By acquiring a plurality of forward training samples, a classifier can be obtained through training, and users who have the possibility of purchasing and reading can be identified through the classifier, so that accurate file pushing can be realized.

In the actual training process, samples of users who purchase articles and have reading behaviors can be used as training samples, and some samples without reading behaviors can be randomly selected to be used as negative samples to train the model.

After the model is obtained through training, the model can be applied to online classification recognition, specifically, in the application process, the search behavior data and the reading behavior data of the target user, the personal information of the user and the like can be input into the classification model, and a classification result is finally obtained, that is, whether the target user has the intention of reading the file is determined, so that whether the file needs to be pushed to the target user is determined.

S4: and (3) pushing the file:

when a target user is determined to search products, the documentations which are possibly interested by the target user can be predicted according to the trained model, when the situation that the pushing condition is met is determined, the documentations can be pushed to the user, the user can display the article sets pushed to the user by clicking the documentations, and each article can be a set of a plurality of commodities, some purchasing guidelines and the like.

At the time of pushing, the frequency of pushing the document to the user may be set, for example, at most once a day, or once a week, etc., so as not to cause the viewing burden of the user.

That is, when the user has a certain amount of search browsing behaviors and no purchasing behaviors occur, whether the user has a strong purchasing content requirement under the leaf category can be predicted in real time through the trained model. The model can be updated in a minute-level quasi-real-time streaming mode, and an online learning mode can be adopted, so that the prediction result is more accurate.

Specifically, as shown in fig. 5, an article may be searched in a content library search engine according to a search word recently input by a user, and a personalized document may be recommended to the user based on a prediction result obtained by the prediction model. When the file is constructed, only leaf categories can be pushed, in order to generate attractive files, the titles of popular articles can be mined, so that a plurality of alternative files are generated, and then the optimal or best file is selected from the candidate files and pushed to a target user.

When the method is realized, an auditing process can be set, and illegal and unhealthy categories can be filtered out, so that the legality of the pushed file can be ensured.

When actual pushing is carried out, for a user with higher content preference, when the user is detected to open a corresponding application, pushed files, abstracts, main pictures and the like can be directly inserted into a product browsing result, and the products can be highlighted. Or sending a recommendation message containing a document through an application message and the like under the condition that the user is detected to quit the application, and realizing article reading by clicking the document by the user.

The leaf categories mentioned above mainly take into account that for the e-commerce platform, the number of products is large, which causes the problem of data sparsity to be aggravated, and the calculation amount required for model training is also large, so in this example, the products are replaced by the leaf categories, wherein the leaf categories are subdivided product categories, and are more or less ten thousand.

When the leaf category characteristics are constructed, the leaf category collaborative matrix of M × M may be constructed by directly using the user purchase behavior data of the leaf categories, where M is the number of the leaf categories, and an element in an ith row and a jth column of the matrix represents the number of users who have browsed and purchased the ith category and the jth category at the same time (for example, simultaneously purchasing a mobile phone and a mobile phone film). Furthermore, dimension reduction processing can be performed on the M x M leaf category collaborative matrix, a main feature vector is extracted, data noise is filtered, and finally a feature with a certain dimension can be retained, for example, a feature with a TOP 200 dimension is retained.

In the implementation process, considering that the offline computation needs a distributed computation framework, and the training model needs to process iterative computation of large-scale data, therefore, a big data computation engine such as Spark or Parameter Server can be adopted.

In the above example, it is considered that in the e-commerce platform, a user sometimes needs to read a relevant shopping guide article when making a shopping decision, so as to further understand the background information, evaluation experience, and the like of a product. In the embodiment, whether the user needs to read the shopping guide article or not is accurately detected when the user wants to purchase a certain product, and the article is accurately put in the user to help the user to make a shopping decision, so that the problem that the behavior data of the user reading the article of the e-commerce platform is less is solved, and the accurate pushing of the file can be realized through the searching and purchasing behavior and a small amount of reading behavior of the user.

In this example, there is also provided a document pushing system, as shown in fig. 6, which may include: the online calculation module, the offline calculation module, the document generation module and the document display module.

The off-line module provides off-line characteristics and trains an off-line model; the online module may comprise two parts: a quasi-real-time user preference prediction module for minute-level updates and an online feature learning module for hour-level updates. The file generation module excavates alternative files from article titles through pattern excavation, and then obtains final files through manual selection. For example, in the case where the "how to choose" and the old driver with you fly ", the blank space is left so that after the" one-piece dress "is recognized, the filling can be performed, and the case" how to choose and the old driver with you fly "is obtained. The presentation module is used for presenting the recommendation result to the user, for example, the recommendation result can be pushed by an application message or inserted into a product search browsing interface.

In the above example, the user content preferences are mined from a very small number of article reading behaviors using the correlation between the user reading articles and searching for purchased products. Mass shopping behavior data owned by e-commerce platforms, such as: the product browsing, collecting and purchasing behaviors of the user and the article reading behaviors of a small number of users recommend documents which may be interested in the shopping process of a large number of users, and the articles in the documents can be read after the user clicks the documents so as to realize shopping decision.

That is, when document pushing is performed, association between shopping behaviors and recommended articles is mainly considered, so that the problem that under the condition that article reading behaviors are deficient, massive user shopping behaviors of a commercial platform are utilized, personalized requirements of users are analyzed, appropriate documents are recommended, articles can be recommended without purchasing the articles by the users, and the actual reading requirements of the users are considered, for example: milk, fruits, toilet paper and the like which are bought by a user every week generally cannot be pushed. Even if different users purchase the same commodity, the same user can not push different documents at different shopping stages, and the fatigue of the users can not be increased by blindly pushing a large number of documents which are not interesting to the users.

The method for pushing the above-mentioned document will be described with reference to another implementation scenario, and in this example, document pushing in a friend circle or a public number is taken as an example.

Specifically, for example, the case push in the friend circle is taken as an example, in order to push the case more accurately, historical behavior data of some users interested in opening the case can be acquired, training or cluster analysis can be performed based on the historical behavior data and the user characteristics of the users, so that the characteristics of the part of people are obtained, target people can be classified based on the historical behavior data and the user characteristics, and accordingly, the users who wish to open the case are identified, and the users who do not wish to open the case are identified, so that the case push effect is improved, the pushed case is opened, and the possibility that promoted articles are purchased is improved.

Further, the groups of people can be further classified, for example, some users like automobile type files, some users like cosmetic type files, some users like mother-baby type files, some users like food type files and the like, when file pushing and the pushed groups are selected, if the file pushing is carried out, the file pushing is preferentially pushed to the users who like the automobile type files, and if the file pushing is a cosmetic type files, the file pushing is preferentially pushed to the users who like the cosmetic type files, so that the possibility that the pushed files are opened and the promoted items are purchased is further improved. .

In the above specific examples, the scenes such as the shopping platform and the friend circle are only taken as examples and are only exemplary illustrations, and the above document pushing method may also be applied to other scenes, for example, document pushing in a public number, document pushing in a news interface, and the like, which is not limited in the present application.

The method embodiments provided in the above description of the present application may be executed in a server, a computer cluster, or a similar computing device. Taking the example of running on a server, fig. 7 is a hardware structure block diagram of the server of the document pushing method according to the embodiment of the present invention. As shown in fig. 7, the server 10 may include one or more (only one shown) processors 102 (the processors 102 may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission module 106 for communication functions. It will be understood by those skilled in the art that the structure shown in fig. 7 is only an illustration and is not intended to limit the structure of the electronic device. For example, the server 10 may also include more or fewer components than shown in FIG. 7, or have a different configuration than shown in FIG. 7.

The memory 104 may be used to store software programs and modules of application software, such as program instructions/modules corresponding to the document pushing method in the embodiment of the present invention, and the processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the document pushing method of the application program. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 104 may further include memory located remotely from processor 102, which may be connected to server 10 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission module 106 is used to receive or transmit data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the server 10. In one example, the transmission module 106 includes a Network adapter (NIC) that can be connected to other Network devices through a base station to communicate with the internet. In one example, the transmission module 106 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

In a software implementation, a request initiating unit, a response receiving unit and a password presenting unit can be included in the obtaining module. Wherein:

the acquisition module is used for acquiring historical search purchasing behavior data of a plurality of users and reading behavior data corresponding to the historical search purchasing behavior data;

the establishing module is used for establishing the relation between the searching purchasing behavior and the reading behavior according to the plurality of purchasing behavior data and the plurality of reading behavior data;

and the pushing module is used for pushing the file to the target user according to the established relation between the searching purchasing behavior and the reading behavior.

In one embodiment, the push module is used for acquiring search behavior data of a target user; acquiring user characteristics of the target user; determining whether to push a file to a user and the reading interest category of the target user according to the established relation between the searching and purchasing behavior and the reading behavior, the searching behavior data of the target user and the user characteristics; and pushing the file related to the reading interest category to the target user.

In one embodiment, the establishing module may be configured to determine whether the user has a reading behavior under the corresponding search purchasing behavior according to the plurality of historical search purchasing behavior data and the reading behavior data corresponding to each historical search purchasing behavior data.

In one embodiment, the pushing module may push the document on a search browsing page of the target user; and/or pushing the file to the target user in a mode of applying information.

In a software implementation, there is provided a document pushing apparatus, which may include: the device comprises an acquisition module, a determination module and a push module, wherein:

the acquisition module can be used for acquiring behavior data and user characteristics of a user;

the determining module is used for determining whether a reading file needs to be pushed to the user or not through a preset classification model;

and the pushing module is used for pushing the reading file to the user under the condition that the swimming suit is determined to need to push the reading file.

In an embodiment, the pushing module may be specifically configured to determine a reading interest category of the target user; and pushing the file related to the reading interest category to the target user. Specifically, the file can be pushed on a browsing page of a target user; and/or pushing the file to the target user in a mode of applying information.

In one embodiment, the pushing module may specifically push the document to the target user in the case that it is determined that the user has a search behavior within a predetermined time period but has no purchase behavior matching the search behavior.

In one embodiment, the user characteristics may include, but are not limited to, at least one of: gender, age, location, purchasing power, category preferences.

In one embodiment, the classification model may be established according to the following steps:

s1: acquiring historical behavior data of a plurality of users and reading behavior data corresponding to the historical behavior data;

s2: extracting user data of a user who reads the file and purchases the file from the historical behavior data and the corresponding reading behavior data as a positive sample;

s3: and training to obtain a classification model through the positive sample, wherein the classification model is used for determining whether the file needs to be pushed to the target user according to the behavior of the target user.

According to the classification model establishing method, the file pushing method and the server, the association degree between the purchasing behavior and the reading behavior of the user is established, so that the group which purchases and is willing to read the file can be effectively determined, and the pushing value and the clicked rate of the pushed file can be improved. By means of the method, the technical problems that an existing file pushing effect is poor and accuracy is low are solved, the technical effect of accurately and efficiently pushing files is achieved, and the file pushing method and the server provided by the application push files to target users by establishing the relation between the search purchasing behaviors and the reading behaviors, so that the pushed files are determined according to the obtained association between a large number of search purchasing behaviors and whether articles are read, and the file pushing accuracy is higher. By means of the mode, the technical problems that an existing file pushing effect is not good and accuracy is low are solved, and the technical effect of accurately and efficiently pushing files is achieved.

Although the present application provides method steps as described in an embodiment or flowchart, additional or fewer steps may be included based on conventional or non-inventive efforts. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of sequences, and does not represent a unique order of performance. When an actual apparatus or client product executes, it may execute sequentially or in parallel (e.g., in the context of parallel processors or multi-threaded processing) according to the embodiments or methods shown in the figures.

The apparatuses or modules illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. The functionality of the modules may be implemented in the same one or more software and/or hardware implementations of the present application. Of course, a module that implements a certain function may be implemented by a plurality of sub-modules or sub-units in combination.

The methods, apparatus or modules described herein may be implemented in computer readable program code to a controller implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, Application Specific Integrated Circuits (ASICs), programmable logic controllers and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may therefore be considered as a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

Some of the modules in the apparatus described herein may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

From the above description of the embodiments, it is clear to those skilled in the art that the present application can be implemented by software plus necessary hardware. Based on such understanding, the technical solutions of the present application may be embodied in the form of software products or in the implementation process of data migration, which essentially or partially contributes to the prior art. The computer software product may be stored in a storage medium such as ROM/RAM, magnetic disk, optical disk, etc., and includes instructions for causing a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to perform the methods described in the various embodiments or portions of the embodiments of the present application.

The embodiments in the present specification are described in a progressive manner, and the same or similar parts in the embodiments are referred to each other, and each embodiment focuses on differences from other embodiments. All or portions of the present application are operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.

While the present application has been described with examples, those of ordinary skill in the art will appreciate that there are numerous variations and permutations of the present application without departing from the spirit of the application, and it is intended that the appended claims encompass such variations and permutations without departing from the spirit of the application.

Claims

1. A classification model building method, characterized in that the method comprises:

extracting users with purchasing behaviors and reading behaviors from the purchasing behavior data and the reading behavior data as training samples;

Training to obtain a classification model through the training sample, wherein the classification model is used for determining whether reading documentations need to be pushed to a target user according to behavior data of the user;

acquiring behavior data of a target user;

acquiring user characteristics of the target user;

determining whether a reading file needs to be pushed to the target user or not through the classification model;

determining the reading interest category of the target under the condition that the reading file needs to be pushed;

pushing a file related to the reading interest category to the target user;

wherein, the pushing of the documentation related to the reading interest category to the target user comprises: and in the case that the user is determined to have the searching behavior within the preset time period and has no purchasing behavior matched with the searching behavior, pushing the file to the target user.

2. The method of claim 1, wherein pushing a document related to the reading interest category to the target user comprises:

pushing the file on a browsing page of a target user;

and/or the presence of a gas in the gas,

and pushing the file to the target user in an application information mode.

3. The method of claim 1, wherein the user characteristics comprise at least one of: gender, age, location, purchasing power, category preferences.

4. A document pushing method, comprising:

acquiring behavior data and user characteristics of a user;

determining whether a reading file needs to be pushed to the user or not through a preset classification model;

under the condition that the case needs to be pushed, pushing a reading case to the user;

wherein pushing a reading case to the user comprises: and determining the reading interest category of the user, and pushing a file related to the reading interest category to the user under the condition that the user is determined to have a search behavior within a preset time period and have no purchase behavior matched with the search behavior.

5. The method of claim 4, wherein pushing a reading document to the user comprises:

pushing the file on a browsing page of a user;

and/or the presence of a gas in the gas,

and pushing the reading file to the user in an application information mode.

6. The method of claim 4, wherein the user characteristics comprise at least one of: gender, age, location, purchasing power, category preferences.

7. The method of any of claims 4 to 6, wherein presetting the classification model comprises:

8. A server comprising a processor and a memory for storing processor-executable instructions which, when executed by the processor, implement the steps of the method of any one of claims 1 to 3.

9. A terminal device comprising a processor and a memory for storing processor-executable instructions which, when executed by the processor, implement the steps of the method of any one of claims 4 to 7.

10. A computer readable storage medium having stored thereon computer instructions which, when executed, implement the steps of the method of any one of claims 1 to 3.

11. A computer readable storage medium having stored thereon computer instructions which, when executed, implement the steps of the method of any one of claims 4 to 7.