CN106126544B - Internet content delivery method and device - Google Patents
Internet content delivery method and device Download PDFInfo
- Publication number
- CN106126544B CN106126544B CN201610424892.7A CN201610424892A CN106126544B CN 106126544 B CN106126544 B CN 106126544B CN 201610424892 A CN201610424892 A CN 201610424892A CN 106126544 B CN106126544 B CN 106126544B
- Authority
- CN
- China
- Prior art keywords
- access
- access record
- record
- characteristic data
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Theoretical Computer Science (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Economics (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Transfer Between Computers (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A method and a device for delivering Internet contents are provided, the method comprises the following steps: obtaining historical access record data of a current user on a current website, and segmenting the historical access record data into a plurality of continuous access records, wherein each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, and each access record comprises access behavior data of the current user on an object on the current website; selecting positive example samples and negative example samples from access records contained in a plurality of continuous access records; summarizing access characteristic data from the historical access record data according to the positive example sample and the negative example sample; performing model training based on the access characteristic data to obtain a plurality of regression models; determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models. The scheme enables the internet content to be delivered more accurately.
Description
Technical Field
The invention relates to the technical field of internet, in particular to a method and a device for delivering internet content.
Background
Nowadays, with the development of the internet, especially the mobile internet, the information that the internet can provide to users is more and more abundant, and users can pay attention to different types of information contents through the internet. For example, for news content, a user may access sports-type information, science-type information, finance-type information, entertainment-type information, and the like via the internet.
According to the historical behaviors of the user on the website, such as browsing, searching, collecting and the like, related contents of objects, such as commodities and the like, which the user visits can be pushed to the user. In the prior art, a recommendation method usually configures simple processing rules according to services and experience, for example, features selected by the processing rules are a time sequence of objects browsed by a user, a recommended content delivery sequence is arranged in a sequence from near to far from the current time, and the possibility that the user clicks the object browsed most recently is estimated to be higher.
However, the order of preferential clicks of recommended contents by the user predicted according to the method of the related art is not accurate enough, in other words, the contents recommended with higher priority cannot accurately represent the contents clicked by the user with priority.
Disclosure of Invention
The technical problem to be solved by the invention is to provide the internet content delivery method and device, so that the accuracy of internet content delivery is improved.
In order to solve the above technical problem, an embodiment of the present invention provides a method for delivering internet content, where the method includes:
obtaining historical access record data of a current user on a current website, and segmenting the historical access record data into a plurality of continuous access records, wherein each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, and each access record comprises access behavior data of the current user on an object on the current website; selecting positive example samples and negative example samples from the access records contained in the plurality of continuous access records; generalizing access characteristic data from the historical access record data according to the positive case sample and the negative case sample; performing model training based on the access characteristic data to obtain a plurality of regression models; determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models.
Optionally, the selecting positive and negative examples from the access records included in the plurality of persistent access records comprises:
for an access record in each persistent access record, marking the access record as the positive example if the object for which the access record is intended was accessed in a previous persistent access record;
for an access record in each persistent access record, if the object for which the access record is intended was accessed in a previous persistent access record but was not accessed in the current persistent access record, marking the access record as the negative example.
Optionally, the regression model is a GBDT tree model.
Optionally, determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models includes:
obtaining the score of the object to be predicted based on the updated access characteristic data of the object to be predicted and the regression models;
and sorting the objects from high to low according to the scores, and determining the delivery priority of the recommended contents associated with the objects to be predicted according to the sorting.
Optionally, the access characteristic data includes characteristic data of a current user accessing the current website, characteristic data of an object being accessed, and characteristic data of the current user accessing the object.
An embodiment of the present invention further provides a device for delivering internet content, where the device includes:
the acquisition unit is suitable for acquiring historical access record data of a current user on a current website and segmenting the historical access record data into a plurality of continuous access records, each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, and each access record comprises access behavior data of the current user on an object on the current website;
the selecting unit is suitable for selecting positive example samples and negative example samples from the access records contained in the plurality of continuous access records;
an induction unit adapted to induce access profile data from the historical access record data according to the positive examples and the negative examples;
the model training unit is suitable for carrying out model training on the basis of the access characteristic data to obtain a plurality of regression models;
a determination unit adapted to determine an order of delivery priority of recommended content associated with a plurality of objects to be predicted based on the plurality of regression models and updated access characteristic data of the plurality of objects to be predicted.
Optionally, the selecting unit includes:
a first marking subunit adapted to, for an access record in each persistent access record, mark the access record as the positive example if the object for which the access record is intended was accessed in a previous persistent access record;
a second marking subunit adapted to mark, for an access record in each persistent access record, an access record as the negative example if the object for which the access record is intended was accessed in a previous persistent access record but not accessed in the current persistent access record.
Optionally, the regression model is a GBDT model.
Optionally, the determining unit includes:
the score obtaining subunit is suitable for obtaining the score of the object to be predicted based on the updated access characteristic data of the object to be predicted and the regression models;
and the delivery order determining subunit is suitable for sorting according to the scores from high to low and determining the delivery priority order of the recommended contents associated with the plurality of objects to be predicted according to the sorting.
Optionally, the access characteristic data includes characteristic data of a current user accessing the current website, characteristic data of an object being accessed, and characteristic data of the current user accessing the object.
Compared with the prior art, the technical scheme of the embodiment of the invention has the following beneficial effects:
according to the technical scheme, historical access record data of a current user on a current website are obtained, the historical access record data are divided into a plurality of continuous access records, positive examples and negative examples are selected from the access records contained in the continuous access records, access characteristic data are summarized for each positive example and each negative example, model training is performed on the basis of the access characteristic data marked as positive examples or negative examples to obtain a plurality of regression models, and then the release priority sequence of recommended contents associated with the objects to be predicted is determined through the regression models and the updated access characteristic data of the objects to be predicted. In the process, the historical access records of the user are divided into a plurality of continuous access records according to the continuous access behaviors of the user, the continuous access records are used as a judgment reference, positive samples and negative samples are selected from the access records contained in the continuous access records, the positive samples and the negative samples provide more accurate model training optimization targets, the regression model obtained based on the target training can more accurately sort the objects to be predicted, the sorting can more accurately represent the probability sequence of clicking the objects to be predicted by the user, and when the priority sequence of the recommended contents delivered to the user is determined according to the sequence, the probability of clicking the recommended contents with higher current priority by the user is higher, so that the accuracy of delivering the recommended contents is improved. Because the delivered recommended content is more accurate, the operations of searching and browsing for many times for obtaining the content which the user is interested in can be better avoided, and further network system resources required for responding to the re-access or the search of the user can be saved. Meanwhile, the possibility that the user acquires the interested content from the recommended content is increased, so that the possibility that the user needs to search, browse and other operations for many times is reduced, and the user experience is improved.
Drawings
Fig. 1 is a flowchart of a method for delivering internet content according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a method for delivering internet content according to an embodiment of the present invention.
Detailed Description
As mentioned in the background, the prior art method may not predict the preferred click sequence of the recommended content by the user accurately, in other words, the preferred recommended content may not represent the preferred click content of the user accurately. Therefore, the user often needs to perform operations such as multiple searches and browsing to obtain the content of interest, and more network system resources need to be provided to respond to the user's browsing or searching again, which leads to an increase in cost. Meanwhile, the user experience is also poor.
According to the technical scheme, historical access record data of a current user on a current website are obtained, the historical access record data are divided into a plurality of continuous access records, positive examples and negative examples are selected from the access records contained in the continuous access records, access characteristic data are summarized for each positive example and each negative example, model training is performed on the basis of the access characteristic data marked as positive examples or negative examples to obtain a plurality of regression models, and then the release priority sequence of recommended contents associated with the objects to be predicted is determined through the regression models and the updated access characteristic data of the objects to be predicted. In the process, the historical access records of the user are divided into a plurality of continuous access records according to the continuous access behaviors of the user, the continuous access records are used as a judgment reference, positive samples and negative samples are selected from the access records contained in the continuous access records, the positive samples and the negative samples provide more accurate model training optimization targets, the regression model obtained based on the target training can more accurately sort the objects to be predicted, the sorting can more accurately represent the probability sequence of clicking the objects to be predicted by the user, and when the priority sequence of the recommended contents delivered to the user is determined according to the sequence, the probability of clicking the recommended contents with higher current priority by the user is higher, so that the accuracy of delivering the recommended contents is improved. Because the delivered recommended content is more accurate, the operations of searching and browsing for many times for obtaining the content which the user is interested in can be better avoided, and further network system resources required for responding to the re-access or the search of the user can be saved. Meanwhile, the possibility that the user acquires the interested content from the recommended content is increased, so that the possibility that the user needs to search, browse and other operations for many times is reduced, and the user experience is improved.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in detail below.
Fig. 1 is a flowchart of a method for delivering internet content according to an embodiment of the present invention. The following is described with reference to the steps shown in fig. 1.
Step S101: obtaining historical access record data of a current user on a current website, and segmenting the historical access record data into a plurality of continuous access records, wherein each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, and each access record comprises access behavior data of the current user on an object on the current website.
The current website is a website to which internet content is to be delivered, and may be any appropriate website, such as a news website, a video website, a shopping website, and the like.
The historical access record data of the current website is from the server of the current website, and the historical access record data may be limited by a time range before a certain time point, for example, the access behavior data of the previous month is taken as the historical access record data in this embodiment.
In a specific implementation, the persistent access record includes at least one access record formed by consecutive access behaviors of the user. The continuous access behavior may be an access behavior in a session (session) process, for example, a process from opening the website S to closing the website S by the user U is a session process of the user U on the website S. Each access record comprises access behavior data of the current user to an object on the current website. Specifically, each access record may include access behavior data consisting of access time, access object, whether it is the first access of the current session, and access type. The "object" referred to by the embodiments of the present invention may be a news category, a video category, a commodity, or other specific object. The access types include browsing and searching. When the object is a commodity, the access type may further include collecting and joining a shopping cart. The following description takes the partial access behavior data of the user U in the website S in table 1 as an example:
TABLE 1
In table 1, the access behavior data of the session defined by Ts1 to Ts3 is the first persistent access record S1, which includes 3 access records. The session access behavior data delimited by Ts4 and Ts5 is a second persistent access record S2, comprising 2 access records. Each access record includes the time of access, whether it is the first access within the current session, the goods accessed, and the type of access. It should be noted that the access behavior data is not limited to the above list.
Step S102: and selecting positive example samples and negative example samples from the access records contained in the plurality of continuous access records.
In a specific implementation, for an access record in each persistent access record, if an object for which the access record is intended has been accessed in a previous persistent access record, marking the access record as the positive example; for an access record in each persistent access record, if the object for which the access record is intended was accessed in a previous persistent access record but was not accessed in the current persistent access record, marking the access record as the negative example.
Continuing with Table 1 to illustrate the selection of the positive and negative examples, it is assumed that the first continuous access record is the first continuous access record in the selected historical access record data, in other words, the historical access record data begins at time Ts 1. Wherein:
for the first continuous access record S1, any merchandise has not been accessed before, and therefore no positive and negative examples are generated in the first continuous access record S1. For the second persistent access record S2, since the article Pid1 was accessed in the first access record S1 and also accessed in the current persistent access record S2, the access record corresponding to the article Pid1 is marked as a positive sample; the good Pid2 was accessed in the first persistent access record S1, but not in the current second persistent access record S2, and therefore this access record corresponding to the good Pid2 is marked as a negative sample.
Similarly, positive and negative examples can be selected for other persistent access records in the selected historical access record data.
It should be noted that, when marking the positive example and the negative example, it may be conditioned not only by being browsed or searched, but also by restricting the preset access type to be satisfied. For example, when the object is a commodity and the access type includes browsing, searching, collecting and adding to a shopping cart, the access record before the current continuous access record and the object whose access type is the preset type in the current continuous access record are marked as the positive example, the access type in the access record before the current continuous access record is the preset type, but the object whose access type is not the preset type in the current continuous access record is marked as the negative example, and the preset type includes searching and browsing. When selecting positive and negative examples, samples considered to be valuable can be selected as a set of model training by specifying preset types to be satisfied.
Step S103: generalizing access profile data from the historical access record data based on the positive examples and the negative examples.
In a specific implementation, the access characteristic data includes characteristic data of the current website accessed by the current user, characteristic data of the object accessed by the current user, and characteristic data of the object accessed by the current user. The feature data of the current website accessed by the current user may include statistical data of an access behavior of the current website accessed by the current user, for example, the feature data may be times of accessing the current website by the current user in a preset period, times of browsing different objects, times of searching different objects, and the like; the accessed characteristic data of the object can comprise statistical data of different objects, such as the number of times that different objects are browsed in each continuous access record, the number of times that different objects are searched, and the like; the characteristic data of the current user accessing the object may be statistical data of the current user's access behavior to different objects, for example, the current user may browse times, search times, last access time, and the like, for different objects in each continuous access record.
Taking the positive example and the negative example selected from table 1 in step S102 as an example, the object is still a commodity.
The feature data of the current user's access to the current website may include: before Ts4, the total number of times the current user U visits the current website S is 3, the last visit time of the current user U is Ts3, and so on.
The accessed characteristic data of the object may include: based on the product Pid1 statistics, the product Pid1 was accessed a total of 2 times (i.e., 1 search and 1 browse) before Ts 4; based on the good Pid2 statistics, the good Pid2 was visited a total of 1 time (i.e., 1 search) before Ts 4.
The accessing of the feature data of the object by the current user U may include: before Ts4, the current user U visits merchandise Pid1 2 times (i.e., searches 1 time and browses 1 time); before Ts4, the current user U visits item Pid2 1 times (i.e., searches 1 time); before Ts4, the last visit time of the current user U to the commercial article Pid1 is Ts3, and the last visit time of the current user U to the commercial article Pid2 is Ts 2.
It should be noted that the access characteristic data may also be summarized according to other preset criteria, which is not limited herein.
Step S104: and performing model training based on the access characteristic data to obtain a plurality of regression models.
In a specific implementation, the regression model may be a Gradient Boosting Decision Tree (GBDT) model. The process of model training is well known to those skilled in the art and will not be described herein.
Step S105: determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models.
The object to be predicted is a candidate object which is possibly subjected to content recommendation aiming at the current user. For example, recommended content associated with 8 objects needs to be selected from 100 candidate objects to be delivered to the user, where the 100 candidate objects are objects to be predicted.
In specific implementation, the scores of the objects to be predicted can be obtained based on the updated access characteristic data of the objects to be predicted and the multiple regression models, the objects to be predicted are ranked from high to low according to the scores, and the delivery priority order of the recommended content associated with the multiple objects to be predicted is determined according to the ranking. For example, the objects to be predicted include a first object and a second object, the prediction is performed according to the access characteristic data updated by the first object and the second object and the regression models, and if the score of the first object is higher than that of the second object, the recommended content of the first object is preferentially delivered to the current user.
In specific implementation, recommended content associated with an object to be predicted with the highest placement priority order may be selected for placement, or recommended content associated with an object to be predicted with the top N placement priority orders may be selected for placement. For example, based on the updated access characteristic data of 100 objects to be predicted and the multiple regression models, the score of each object to be predicted may be obtained, and according to the order of the scores from high to low, the recommended content associated with the object to be predicted with the highest score is selected to be delivered to the user, or the recommended content associated with the object to be predicted with the top 8 scores is selected to be delivered to the user. For example, when the object is a news category, the recommended content may be news content; when the object is a commodity, the recommended content may be an internet advertisement.
It should be noted that, in steps S101 to S105 of the embodiment of the present invention, for a specific current user, the same implementation manner may be used for other users accessing the current website to determine the delivery priority of the recommended content for the user.
In the embodiment of the invention, the positive sample and the negative sample selected from the access records contained in the continuous access records provide the target for model training optimization, and the regression model obtained based on the target training can perform more accurate sequencing on the objects to be predicted, so that the sequencing can more accurately represent the possibility sequence of the user clicking the objects to be predicted, further the accuracy of estimating the possibility of the user clicking the recommended content is improved, when the priority sequence of the recommended content released to the user is determined according to the sequence, the possibility of the user clicking the recommended content with higher priority is higher, thus the user is better prevented from performing operations such as multiple searching and browsing for obtaining the content of interest of the user, and the network system resources required for responding to the user to revisit or search for the content of interest are saved. Meanwhile, the possibility that the user acquires the interested content from the recommended content is increased, so that the possibility that the user needs to search, browse and other operations for many times is reduced, and the user experience is improved.
Fig. 2 is a schematic structural diagram of an internet content delivery apparatus in an embodiment of the present invention. The internet content delivery apparatus shown in fig. 2 may include: an acquisition unit 201, a selection unit 202, a generalization unit 203, a model training unit 204, and a determination unit 205.
The obtaining unit 201 is adapted to obtain historical access record data of a current website of a current user, and segment the historical access record data into a plurality of continuous access records, where each continuous access record includes at least one access record formed by continuous access behaviors of the current user to the current website, and each access record includes access behavior data of an object on the current website of the current user.
The selecting unit 202 is adapted to select a positive example and a negative example from the access records included in the plurality of persistent access records.
In a specific implementation, the selecting unit 202 may include:
a first marking subunit adapted to, for an access record in each persistent access record, mark the access record as the positive example if the object for which the access record is intended was accessed in a previous persistent access record;
a second marking subunit adapted to mark, for an access record in each persistent access record, an access record as the negative example if the object for which the access record is intended was accessed in a previous persistent access record but not accessed in the current persistent access record.
The induction unit 203 is adapted to induce access characteristic data from the historical access record data according to the positive examples and the negative examples.
The model training unit 204 is adapted to perform model training based on the access characteristic data to obtain a plurality of regression models.
In a specific implementation, the regression model may be a GBDT model.
The determining unit 205 is adapted to determine an order of delivery priority of recommended content associated with a plurality of objects to be predicted based on the plurality of regression models and the updated access characteristic data of the plurality of objects to be predicted.
In a specific implementation, the determining unit 205 may include:
the score obtaining subunit is suitable for obtaining the score of the object to be predicted based on the updated access characteristic data of the object to be predicted and the regression models;
and the delivery order determining subunit is suitable for sorting according to the scores from high to low so as to determine the delivery priority order of the recommended contents associated with the plurality of objects to be predicted according to the sorting.
The structural description and advantageous effects of the internet content delivery device may correspond to the implementation description and advantageous effects of the internet content delivery method described with reference to fig. 1, and are not described herein again.
In a specific implementation, when the internet content delivery method can be applied to the field of internet advertisements, the internet content delivery device can be applied to a DSP server.
Those skilled in the art will appreciate that all or part of the steps in the methods of the above embodiments may be implemented by instructions associated with hardware via a program, which may be stored in a computer-readable storage medium, and the storage medium may include: ROM, RAM, magnetic or optical disks, and the like.
The method and system of the embodiments of the present invention have been described in detail, but the present invention is not limited thereto. Various changes and modifications may be effected therein by one skilled in the art without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (10)
1. A method for delivering internet content, comprising:
obtaining historical access record data of a current user on a current website, and segmenting the historical access record data into a plurality of continuous access records, wherein each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, each access record comprises access behavior data of the current user on an object on the current website, and each continuous access behavior is an access behavior in a session process;
selecting positive example samples and negative example samples from the access records contained in the plurality of continuous access records;
generalizing access characteristic data from the historical access record data according to the positive case sample and the negative case sample;
performing model training based on the access characteristic data to obtain a plurality of regression models;
determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models.
2. The method of delivering internet content according to claim 1, wherein the selecting positive examples and negative examples from the access records included in the plurality of persistent access records comprises:
for an access record in each persistent access record, marking the access record as the positive example if the object for which the access record is intended was accessed in a previous persistent access record;
for an access record in each persistent access record, if the object for which the access record is intended was accessed in a previous persistent access record but was not accessed in the current persistent access record, marking the access record as the negative example.
3. A method for delivering internet content as claimed in claim 1, wherein said regression model is a GBDT tree model.
4. The internet content delivery method according to claim 1, wherein determining a delivery priority of recommended content associated with a plurality of objects to be predicted based on the updated access characteristic data of the plurality of objects to be predicted and the plurality of regression models comprises:
obtaining the score of the object to be predicted based on the updated access characteristic data of the object to be predicted and the regression models;
and sorting the objects from high to low according to the scores, and determining the delivery priority of the recommended contents associated with the objects to be predicted according to the sorting.
5. The method for delivering internet content according to claim 1, wherein the access characteristic data comprises characteristic data of a current user accessing the current website, characteristic data of an object being accessed, and characteristic data of the current user accessing the object.
6. An internet content delivery apparatus, comprising:
the acquisition unit is suitable for acquiring historical access record data of a current user on a current website and segmenting the historical access record data into a plurality of continuous access records, each continuous access record comprises at least one access record formed by continuous access behaviors of the current user on the current website, each access record comprises access behavior data of the current user on an object on the current website, and each continuous access behavior is an access behavior in a session process;
the selecting unit is suitable for selecting positive example samples and negative example samples from the access records contained in the plurality of continuous access records;
an induction unit adapted to induce access profile data from the historical access record data according to the positive examples and the negative examples;
the model training unit is suitable for carrying out model training on the basis of the access characteristic data to obtain a plurality of regression models;
a determination unit adapted to determine an order of delivery priority of recommended content associated with a plurality of objects to be predicted based on the plurality of regression models and updated access characteristic data of the plurality of objects to be predicted.
7. The internet content delivery apparatus according to claim 6, wherein the selecting unit comprises:
a first marking subunit adapted to, for an access record in each persistent access record, mark the access record as the positive example if the object for which the access record is intended was accessed in a previous persistent access record;
a second marking subunit adapted to mark, for an access record in each persistent access record, an access record as the negative example if the object for which the access record is intended was accessed in a previous persistent access record but not accessed in the current persistent access record.
8. Delivery device of internet content according to claim 6, characterized in that said regression model is the GBDT model.
9. The internet content delivery apparatus according to claim 6, wherein the determining unit comprises:
the score obtaining subunit is suitable for obtaining the score of the object to be predicted based on the updated access characteristic data of the object to be predicted and the regression models;
and the delivery order determining subunit is suitable for sorting according to the scores from high to low and determining the delivery priority order of the recommended contents associated with the plurality of objects to be predicted according to the sorting.
10. The internet content delivery apparatus according to claim 6, wherein the access characteristic data includes a characteristic data of a current user accessing the current website, a characteristic data of an object being accessed, and a characteristic data of the current user accessing the object.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610424892.7A CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610424892.7A CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106126544A CN106126544A (en) | 2016-11-16 |
| CN106126544B true CN106126544B (en) | 2020-01-24 |
Family
ID=57469461
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610424892.7A Active CN106126544B (en) | 2016-06-15 | 2016-06-15 | Internet content delivery method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106126544B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108280670B (en) | 2017-01-06 | 2022-06-21 | 腾讯科技（深圳）有限公司 | Seed crowd diffusion method and device and information delivery system |
| CN107153684B (en) * | 2017-04-24 | 2020-02-07 | 北京小米移动软件有限公司 | Display method, device and equipment of push message |
| CN107633326A (en) * | 2017-09-14 | 2018-01-26 | 北京拉勾科技有限公司 | A kind of user delivers the construction method and computing device of wish model |
| CN108304853B (en) * | 2017-10-10 | 2022-11-08 | 腾讯科技（深圳）有限公司 | Game correlation obtaining method and device, storage medium and electronic device |
| CN110674434B (en) * | 2019-09-26 | 2022-03-29 | 秒针信息技术有限公司 | Method and device for releasing browsing resources |
| CN114417817B (en) * | 2021-12-30 | 2023-05-16 | 中国电信股份有限公司 | Session information cutting method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8175377B2 (en) * | 2009-06-30 | 2012-05-08 | Xerox Corporation | Method and system for training classification and extraction engine in an imaging solution |
| CN105631538A (en) * | 2015-12-23 | 2016-06-01 | 北京奇虎科技有限公司 | User activity prediction method and device, and application method and system thereof |
-
2016
- 2016-06-15 CN CN201610424892.7A patent/CN106126544B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870972A (en) * | 2012-12-07 | 2014-06-18 | 盛乐信息技术（上海）有限公司 | Data recommendation method and data recommendation system |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106126544A (en) | 2016-11-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106126544B (en) | Internet content delivery method and device | |
| CN110020094B (en) | Display method and related device for search results | |
| CN105989004B (en) | Information delivery preprocessing method and device | |
| US10572565B2 (en) | User behavior models based on source domain | |
| JP5721818B2 (en) | Use of model information group in search | |
| WO2017121251A1 (en) | Information push method and device | |
| CN107888950A (en) | A kind of method and system for recommending video | |
| US11836778B2 (en) | Product and content association | |
| CN105718184A (en) | Data processing method and apparatus | |
| US20170154116A1 (en) | Method and system for recommending contents based on social network | |
| CN109753601B (en) | Method and device for determining click rate of recommended information and electronic equipment | |
| WO2015120798A1 (en) | Method for processing network media information and related system | |
| CN110175895B (en) | Article recommendation method and device | |
| CN104462336A (en) | Information pushing method and device | |
| CN108537596B (en) | Method, device and system for recommending vehicle type in search box and memory | |
| US20180285331A1 (en) | Method, server, browser, and system for recommending text information | |
| US9330071B1 (en) | Tag merging | |
| JP6728178B2 (en) | Method and apparatus for processing search data | |
| CN111061954B (en) | Search result sorting method and device and storage medium | |
| CN106033455B (en) | Method and equipment for processing user operation information | |
| CN111400586A (en) | Group display method, terminal, server, system and storage medium | |
| CN113343095A (en) | Model training and information recommendation method and device | |
| US8745042B2 (en) | Determining matching degrees between information categories and displayed information | |
| CN104572887A (en) | Method and system for retrieving product information | |
| CN110020209B (en) | Method and system for determining correlation between content and search word and method and system for displaying correlation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |