CN106503025B

CN106503025B - Application recommendation method and system

Info

Publication number: CN106503025B
Application number: CN201510568010.XA
Authority: CN
Inventors: 肖镜辉; 田乐逍
Original assignee: Beijing Sogou Technology Development Co Ltd
Current assignee: Beijing Sogou Technology Development Co Ltd
Priority date: 2015-09-08
Filing date: 2015-09-08
Publication date: 2021-02-12
Anticipated expiration: 2035-09-08
Also published as: CN106503025A

Abstract

The embodiment of the invention provides an application recommendation method and an application recommendation system, and relates to the technical field of computers. The method comprises the following steps: acquiring user behavior records related to applications and analyzing each application to obtain a first application alternative set; obtaining a recommended application set according to the first application alternative set; and displaying each application in the recommended application set. The embodiment of the invention integrates the access behaviors of the user, obtains a recommendation list of APP with higher quality through relatively objective statistics, reduces the labor cost and improves the recommendation efficiency.

Description

Application recommendation method and system

Technical Field

The present invention relates to the field of computer technologies, and in particular, to an application recommendation method and an application recommendation system.

Background

Mobile terminals represented by mobile phones and tablet computers are developed rapidly in recent years, the performance of a CPU and a storage device is greatly improved, and accordingly, applications on the mobile terminals are more and more abundant, and user interfaces are more and more friendly.

Many functions of a mobile terminal need to be completed by an Application (APP) installed on the mobile terminal. How to find suitable software and find interesting games is a problem for users of mobile terminals when using the mobile terminals. Thus, various application management platforms come into play, such as: the mobile phone assistant for dog searching, the mobile phone assistant for 360 degrees, the mobile phone assistant for 91 degrees, the mobile phone assistant for pea pod, the millet application market, the apple APP STORE and the like. The application management platform aims to help a mobile terminal user to find, search, download, install and update the APP more conveniently. These application management platforms can be installed on the PC or MAC, or directly on the mobile terminal. The application management platform can show organized APP for a user after being started, a secondary APP recommendation page is arranged in the application management platform to show or recommend new on-shelf or high-quality APP for the user, the display content of the APP recommendation page and the display sequence among the APPs directly influence whether the user can quickly find the required high-quality APP on the page, and further influence the direct experience of the user using the application management platform and the download amount, the flow amount and the commercial value of the APP on the application management platform.

However, in the prior art, the APPs and their sequences displayed in the APP recommendation page mainly depend on manual editing, auditing and sequencing. For example, 50 good-quality APPs are displayed in the APP recommendation page, and the APPs need to be screened out from 30 ten thousand APPs in the APP library one by one; and 50 good APPs are ordered in order of high quality to low quality. If the operation is completely done manually, a large amount of human resources are consumed every day, and the efficiency is very low. And the modes of manual editing, auditing and sequencing have strong subjectivity, and the recommended APP cannot be objectively ensured to be the high-quality APP which is used by the user.

Disclosure of Invention

In view of the above, embodiments of the present invention are proposed to provide an application recommendation method and a corresponding application recommendation system that overcome or at least partially solve the above problems.

In order to solve the above problems, the present invention discloses an application recommendation method, including:

acquiring user behavior records related to applications and analyzing each application to obtain a first application alternative set;

obtaining a recommended application set according to the first application alternative set;

and displaying each application in the recommended application set.

Preferably, the step of obtaining a user behavior record related to the application and analyzing each application to obtain a first application candidate set includes:

for each application, extracting a characteristic value of at least one dimension from a user behavior record related to the application; the characteristic values of the at least one dimension include: the characteristic value of a first value dimension and the characteristic value of a user access behavior dimension;

calculating a total characteristic value of the application according to the characteristic value of the extracted at least one dimension and the weight of the corresponding dimension;

ordering the total eigenvalues by N₁Each application joins the first set of application alternatives.

Preferably, for each application, the step of extracting feature values of at least one dimension from a user behavior record associated with the application comprises:

for each application, calculating a ratio between first value data of the application and a distribution amount of the application, the ratio being a feature value of a first value dimension.

Preferably, the characteristic value of the user access behavior dimension includes: browsing the characteristic value of the dimension, and downloading the characteristic value of the dimension.

Preferably, when the feature value of the user access behavior dimension is a feature value of a browsing dimension, the step of extracting, for each application, a feature value of at least one dimension from a user behavior record associated with the application includes:

counting, for each application, a first browsing frequency of the application browsed in an application summary display page, a second browsing frequency of the application browsed in a search result page, and a third browsing frequency of the application browsed in an application detail display page;

and respectively combining the first browsing times, the second browsing times and the third browsing times with the weight of the page to which the browsing times are located, and calculating the characteristic value of the browsing dimensionality.

Preferably, when the feature value of the user access behavior dimension is a feature value of a download dimension, the step of extracting, for each application, a feature value of at least one dimension from a user behavior record associated with the application includes:

counting, for each application, a first download time of the application summary presentation page, a second download time of the application summary presentation page, and a third download time of the application detail presentation page;

and respectively combining the first downloading times, the second downloading times and the third downloading times with the weight of the page where the first downloading times, the second downloading times and the third downloading times, and calculating the characteristic value of the downloading dimension.

Preferably, before the step of obtaining the recommended application set according to the first application candidate set, the method further includes:

acquiring applications in an application recommendation page of at least one application management platform and analyzing each application to obtain a second application alternative set;

according to the first application alternative set, the step of obtaining the recommended application set comprises the following steps:

and fusing the first application alternative set and the second application alternative set to obtain a recommended application set.

Preferably, the step of obtaining applications in the application recommendation page of at least one application management platform and analyzing each application to obtain the second application candidate set includes:

acquiring a corresponding application ranking list for an application recommendation page of at least one application management platform;

when the application ordered list is two or more, performing fusion calculation on each application ordered list to obtain a product containing N₂A second set of application alternatives for each application, wherein N₂Is less than or equal to the sum of the number of applications in all the application lists.

Preferably, when the application ordered lists are two or more, the fusion calculation is carried out on each application ordered list to obtain the result containing N₂The step of the second set of application alternatives for each application comprises:

selecting P applications with the top rank from each application ranking list;

calculating the total characteristic value of the application according to the characteristic value of each dimension and the weight of the corresponding dimension;

ordering the total eigenvalues by N₂The applications join a second set of application alternatives.

Preferably, the step of fusing the first application candidate set and the second application candidate set to obtain the recommended application set includes:

top M of the set of applications will be recommended₁Selecting the top M from the second application candidate set₁Filling by each application;

will recommend the rest of the application set₂Selecting the corresponding M in the first application candidate set₂Filling by the application of the user; wherein M is₁And M₂The sum ofThe number of applications in the set of applications is recommended.

Preferably, after obtaining the recommended application set, the method further includes:

for each application in the recommended application set, labeling the application with a classification label according to a classification label model; and/or

And analyzing the use behaviors of the functions of the application according to the users to construct an application classification label model.

The invention also provides an application recommendation system, comprising:

the first application alternative set generation module is used for acquiring user behavior records related to the applications and analyzing each application to acquire a first application alternative set;

a recommended application set acquisition module, configured to acquire a recommended application set according to the first application alternative set;

and the display module is used for displaying each application in the recommended application set.

The embodiment of the invention has the following advantages:

according to the embodiment of the invention, the quality of the APP is comprehensively evaluated through large-scale analysis of the user behavior records aiming at the APP display page, so that a first application alternative set is constructed. Then, the applications are screened from the first application alternative set according to the requirements of the recommended application set, the applications meeting the requirements are added into the recommended application set, and then the applications in the recommended application set can be displayed in an application recommendation page of the application management platform. The embodiment of the invention integrates the access behaviors of the user, obtains a recommendation list of APP with higher quality through relatively objective statistics, reduces the labor cost and improves the recommendation efficiency.

Drawings

FIG. 1 is a flow chart of the steps of an embodiment of an application recommendation method of the present invention;

FIGS. 1A, 1B, 1C are an APP demonstration example of the present invention;

FIG. 2 is a flow chart of steps in another embodiment of an application recommendation method of the present invention;

FIG. 3 is a flow chart of steps in another embodiment of an application recommendation method of the present invention;

FIG. 4 is a block diagram of an embodiment of an application recommendation system of the present invention;

FIG. 5 is a block diagram of another embodiment of an application recommendation system according to the present invention

Fig. 6 is a block diagram of another embodiment of the application recommendation system of the present invention.

Detailed Description

In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.

One of the core ideas of the embodiment of the invention is that the quality of the APP is comprehensively evaluated through large-scale analysis of user behavior records aiming at the APP display page, so that a first application alternative set is constructed. Then, the applications can be screened from the first application alternative set according to the requirements of the recommended application set, the applications meeting the requirements are added into the recommended application set, and then the applications in the recommended application set can be displayed in an application recommendation page of the application management platform. The embodiment of the invention can integrate the access behaviors of the user, such as behaviors of browsing and downloading applications in different types of pages, relatively objectively count to obtain a recommendation list of APP with higher quality, reduce labor cost and improve recommendation efficiency.

Example one

Referring to fig. 1, a flowchart illustrating steps of an embodiment of an application recommendation method according to the present invention is shown, which may specifically include the following steps:

step 110, obtaining user behavior records related to the applications and analyzing the applications to obtain a first application alternative set.

In the embodiment of the invention, the large-scale user behavior records related to the application can be used, such as the browsing times of a user on a page where the application is located, the downloading times of the application and the like.

It can be understood that the embodiment of the present invention may obtain the user behavior record from the server of each website. Generally, the user behavior record is stored in an access log of the server, and the embodiment of the present invention may obtain the user behavior record related to the application through the access log.

And analyzing the user behavior records to obtain high-quality APPs, and then adding the high-quality APPs into the first application alternative set. In the embodiment of the invention, the high-quality APP can be measured from multiple sides, such as browsing times, downloading times, income brought to an application management platform and the like. Higher number of the above aspects indicates better quality of the APP.

When the embodiment of the invention is applied to an application management platform, and a large number of APPs exist in the application management platform, when a recommended application set is generated, the APPs in the application management platform need to be recommended, so that when user behavior records related to the applications are obtained and the applications are analyzed, the applications in the specified application set are analyzed. The specified application set may include all applications in the application management platform, or may include part of applications in the application management platform, such as applications newly added to the application management platform in the last period of time.

Optionally, step 110 includes:

sub-step a10, for each application, extracting a characteristic value of at least one dimension from a user behavior record associated with said application; the characteristic values of the at least one dimension include: the characteristic value of the first value dimension and the characteristic value of the user access behavior dimension.

Wherein the user access behavior dimension characteristic value comprises: browsing the characteristic value of the dimension, and downloading the characteristic value of the dimension. The characteristic value of the browsing dimension can be understood as a measure of the browsing amount of a certain APP shown in each webpage by the user. The characteristic value of the download dimension can be understood as a measure of the download amount of a certain APP by a user.

In the embodiment of the present invention, feature values of multiple dimensions, such as a first-value dimension, a browsing dimension, and a downloading dimension, may be extracted from the user behavior record. The characteristic value of the first value dimension may be a profit brought by a certain APP unit distribution amount. The characteristic value of the browsing dimension can be understood as a measure of the browsing amount of an APP displayed in each webpage by the user. The characteristic value of the download dimension can be understood as a measure of the download amount of an APP by a user. The dimension characteristic value of the user access behavior can be understood as the flow generated by the user access behavior, such as the browsing amount, the downloading amount, and the like.

Optionally, the sub-step a10 includes:

the substep a101 calculates, for each application, a ratio between the first value data of the application and the distribution amount of the application, and takes the ratio as a characteristic value of the first value dimension.

In the embodiment of the present invention, the first value data may be a business pipeline of an APP, that is, the first value data may be total revenue data obtained by distributing the APP for the application management platform. That is, the APP management platform distributes a large amount of the APPs, and the APP developers pay the APP management platform at a certain price, where the paid price is the business flow or total profit of the APPs under the distribution amount. Then, the first value data of each APP may be obtained, and the distribution amount of each APP in the application management platform may also be obtained, for example, for APPs_IThe APP_IMay be recorded as "APP_ITotal revenue ", the distribution amount of which in the application platform can be recorded as" APP_IDistribution amount ", the characteristic value can be calculated by equation (1):

SCORE_P(APP_I)＝APP_Itotal profit/APP_IDistribution volume (1)

Wherein SCORE_P(APP_I) For a certain APP_IThe characteristic value of (2).

Equation (1) calculates APP_IThe yield brought by unit distribution amount, namely, each time the application management platform distributes one APP_IThe resulting revenue. APP_IThe larger the profit per unit of distribution, the better.

Of course, in the embodiment of the present invention, the total profit of each APP and the distribution amount of each APP in the same time period may be obtained. Such as total revenue per APP for the last two weeks, and the amount of distribution per APP. Therefore, the characteristic value of the first value dimensionality of each APP can be accurately extracted, and the APP can be recommended more accurately.

Optionally, when the characteristic value of the user access behavior dimension is a characteristic value of a browsing dimension, the sub-step a10 includes:

and a substep A102, for each application, counting a first browsing time of the application browsed in the application summary display page, a second browsing time of the application browsed in the search result page and a third browsing time of the application browsing in the application detail display page.

In the embodiment of the invention, the browsing characteristic value of the browsing dimension can be calculated according to the browsing times of the application in various web pages aiming at each application. The various webpages include an APP summary display webpage displaying APP summary information, and the summary information of multiple APPs may be simultaneously displayed in the webpage, as shown in fig. 1A, which is a partial APP display example displayed in the APP summary display page. The above various web pages include APP search result pages, that is, pages obtained by a user searching for an APP in a search page, as shown in fig. 1B, which is a partial example of APP search result pages obtained by a user searching for a "PDF reader" in a search page, from which the user can browse the APP, and certainly, search result pages of other forms may also be included. The various web pages may also include an application detail presentation page, i.e., a page that a user clicks to introduce a specific case of an application, which shows that the user is interested in the application, for example, fig. 1C, which shows an example of a certain application detail presentation page.

For mobile terminals such as mobile phones, when the application management platform application of the dog searching mobile phone assistant is opened on the mobile phone, the application management platform has various channels such as "boutique", "application", "game", "video", "management", "recommendation", "classification", "ranking", "special topic", "information", or various pages, and when any one of the channels is clicked, the channel jumps to the corresponding channel or page, and each channel (page) is an APP display page. The foregoing "TAB page" in the embodiment of the present invention may be understood as the above-described respective channels of the application management platform.

Of course, in the embodiment of the present invention, when the user behavior record is extracted from the user access log of the server, the type of each website accessed by the user is determined. In the embodiment of the invention, the regular expression for judging the website type can be set aiming at the website of each website in advance so as to judge which type each webpage in the website belongs to, thereby carrying out statistics. Wherein, for a regular expression: in the embodiment of the invention, for example, for search pages of various search engines, regular expressions can be set for websites of the search engine pages, and then each regular expression corresponds to a webpage type, and if the regular expressions are matched, it is determined that the websites belong to the webpage type, and if the regular expressions are not matched, the websites do not belong to the webpage type. If the regular expression of the website of the search page of the dog search is \ bSOSO. COM \ b, and the two "\ b" are matched with the so. com character string, if the match indicates that the website is the search page of the dog search, each webpage of the corresponding website is judged to belong to the type of the search result page.

And a substep A103, respectively combining the first browsing times, the second browsing times and the third browsing times with the weight of the page where the browsing times are located, and calculating the characteristic value of the browsing dimension.

In the embodiment of the present invention, different weights may be assigned to each type of page, for example, the weight of the application summary presentation page is Λ 1, the weight of the search result page is Λ 2, and the weight of the application detail presentation page is Λ 3.

The present invention can count the first browsing times C of the application being browsed in the application summary presentation page₁(APP_I) Second browsing times C browsed in search result page₂(APP_I) And a third browsing number C of times of being browsed in the application detail presentation page₃(APP_I)。

Then the feature value of the browsing dimension can be calculated by equation (2):

SCORE_BROWSING(APP_I)＝Λ1*C₁(APP_I)+Λ2*C₂(APP_I)+Λ3*C₃(APP_I) Formula (2)

Wherein SCORE_BROWSING(APP_I) Is APP_IThe characteristic value of the browsing dimension of (1).

Optionally, when the user access behavior dimension characteristic value is a characteristic value of a download dimension, the sub-step a10 includes:

and a substep A104, for each application, counting a first downloading time of the application summary presentation page, a second downloading time of the application to be downloaded in the search result page, and a third downloading time of the application to be downloaded in the application detail presentation page.

In the embodiment of the invention, the browsing characteristic value of the downloading dimension can be calculated according to the downloading channel of each application. The download channel may include: a channel for displaying a webpage through an APP summary displaying the APP summary information, a channel for displaying page downloading through a search result, and a channel for displaying page downloading through application details.

In the embodiment of the present invention, the APP summary display webpage, the application detail display page, and the search result display page may be pages in an application management platform. Then, the downloading channels in other pages, such as channels for downloading other software, for example, the current application management platform is a dog searching mobile phone assistant, and if the user downloads the APP through the APP downloading page of the dog searching input method, the user downloads the APP into other pages.

The various webpages include an APP summary display webpage displaying APP summary information, and the summary information of multiple APPs may be simultaneously displayed in the webpage, as shown in fig. 1A, which is a partial APP display example displayed in the APP summary display page. The above various web pages include APP search result pages, that is, pages obtained by a user searching for an APP in a search page, as shown in fig. 1B, which is a partial example of APP search result pages obtained by a user searching for a "PDF reader" in a search page, from which the user can browse the APP, and certainly, search result pages of other forms may also be included. The various web pages may also include an application detail presentation page, i.e., a page that a user clicks to introduce a specific case of an application, which shows that the user is interested in the application, for example, fig. 1C, which shows an example of a certain application detail presentation page.

And a substep A105, respectively combining the first download times, the second download times and the third download times with the weight of the page where the first download times, the second download times and the third download times are located, and calculating a characteristic value of a download dimension.

In the embodiment of the present invention, different weights may be respectively assigned to the various types of pages, for example, the weight of the application summary presentation page is λ 1, the weight of the search result page is λ 2, and the weight of the application detail presentation page is λ 3.

The present invention can count the first download times D that the application is downloaded in the application summary presentation page and the application detail presentation page₁(APP_I) Second number of downloads D downloaded in search result page₂(APP_I) And a third download number D of times downloaded in being downloaded in the application detail presentation page₃(APP_I)。

Then the characteristic value of the download dimension can be calculated by equation (3):

SCORE_DOWNLOADING(APP_I)＝λ1*D₁(APP_I)+λ2*D₂(APP_I)+λ3*D₃(APP_I) … … formula (3)

Wherein SCORE_DOWNLOADING(APP_I) Is APP_ICharacteristic values of the download dimension.

Of course, the embodiment of the present invention may also obtain the download times of other download channels, for example, the download times of other management distribution platforms to the APP, set a weight for the management distribution platform, multiply the weight by the number of times, and then add the above formula (3).

Sub-step a11, calculating the total eigenvalue of the application according to the eigenvalue of each dimension and the weight of the corresponding dimension.

Then, the invention can calculate the total characteristic value of each APP according to the characteristic value of the first value dimension, the characteristic value of the browsing dimension and the characteristic value of the downloading dimension.

Such as the aforementioned APP_IPractice of the inventionIn an example, a weight may be set for each feature dimension, such as: the weight of the characteristic value of the first value dimension is lambda_PWeight lambda of characteristic value of browsing dimension_LWeight lambda of characteristic value of download dimension_DThen APP_ICan be calculated by equation (4):

SCORE_TOTAL(APP_I)＝λ_P*SCORE_P(APP_I)+λ_L*SCORE_BROWSING(APP_I)+λ_D*SCORE_DOWNLOADING(APP_I) … … formula (4)

Of course, the SCORE can be executed first in the embodiment of the present invention_BROWSINGAnd SCORE_DOWNLOADINGCalculating the first characteristic value SCORE of the overall access behavior of the user together_USERS(APP_I) The first feature value may determine whether the APP is required by most users. Then using SCORE_P(APP_I) And judging whether the APP has greater commercial value. The two are weighted and combined to calculate the total characteristic value SCORE_TOTAL(APP_I)。

Namely, SCORE is calculated by the formula (5)_USERS(APP_I)：

SCORE_USERS(APP_I)＝λ_L1*SCORE_BROWSING(APP_I)+λ_D1*SCORE_DOWNLOADING(APP_I) … … formula (5)

Wherein λ_L1Is SCORE_BROWSING(APP_I) Weight of (2), where λ_D1Is SCORE_DOWNLOADING(APP_I) The weight of (c).

SCORE is then calculated by equation (6)_TOTAL(APP_I)：

SCORE_TOTAL(APP_I)＝λ_P1*SCORE_P(APP_I)+λ_U*SCORE_USERS(APP_I) … … formula (6)

Wherein λ_P1Is SCORE_P(APP_I) Weight of (2), where λ_UIs SCORE_USERS(APP_I) The weight of (c).

Substep A12, sorting the overall eigenvalues by N₁Each application joins the first set of application alternatives.

Then, the SCORE of each APP is obtained by recalculation_TOTAL(APP_I) Then, the N in the top of the sequence can be obtained₁And adding the first application candidate set.

Of course, in the embodiment of the present invention, the SCORE of all APPs in the webpage can be calculated_TOTAL(APP_I) Then, the APP in the APP library of the application management platform is applied according to the SCORE_TOTAL(APP_I) Sorting is performed, and then N in the top sorting is selected₁Each APP joins the first set of application alternatives. N in the front of the sequence₁The APP is the high-quality APP. Wherein N is₁The size of the integer is an integer which is greater than or equal to 0, and the value of the size can be determined according to actual requirements.

And step 120, obtaining a recommended application set according to the first application candidate set.

Then, according to L APPs required by the recommended application set, from the previous N₁And continuously screening L APPs by the APP. Wherein N is₁>L。

The screening being, for example, direct comparison of N₁And adding the top L of the APPs into the recommended application set.

And step 130, displaying each application in the recommended application set.

Then, recommending each application in the application set, that is, putting the relevant information of the application into the corresponding position of the application recommendation page for display according to the sequence, and providing the user with browsing/downloading.

According to the embodiment of the invention, the quality of the APP is comprehensively evaluated through large-scale analysis of the user behavior records aiming at the APP display page, so that a first application alternative set is constructed. Then, the applications can be screened from the first application alternative set according to the requirements of the recommended application set, the applications meeting the requirements are added into the recommended application set, and then the applications in the recommended application set can be displayed in an application recommendation page of the application management platform. The embodiment of the invention integrates the access behaviors of the user, such as behaviors of browsing and downloading applications in different types of pages, relatively objectively counts to obtain a recommendation list of APP with higher quality, reduces the labor cost and improves the recommendation efficiency.

Example two

Referring to fig. 2, a flowchart illustrating steps of an embodiment of an application recommendation method according to the present invention is shown, which may specifically include the following steps:

step 210, obtaining user behavior records related to the applications and analyzing the applications to obtain a first application alternative set.

Step 220, obtaining the applications in the application recommendation page of at least one application management platform and analyzing each application to obtain a second application alternative set.

In practical application, there may also be an application recommendation page or an application recommendation page of an existing application management platform in a network, which is generally manually edited by a technician and has a certain reference value. The method and the system can be used for capturing the specified application recommendation pages of other application management platforms in a specified mode, namely capturing the specified application recommendation pages, analyzing the applications in the specified application recommendation pages, analyzing each application, and obtaining the second application alternative set.

Of course, the embodiment of the present invention may obtain applications in the application recommendation pages of one or more application management platforms.

Optionally, step 220 includes:

and a substep B10, acquiring a corresponding application ranking list for the application recommendation page of at least one application management platform.

In the invention, a webpage grabber can be used for grabbing and analyzing the specified application recommendation page to obtain the application ranking list corresponding to the page.

Substep B11, when the application ordered list is two or more, performing fusion calculation on each application ordered list to obtain a product containing N₂A second set of application alternatives for each application, wherein N₂Is less than or equal to the sum of the number of applications in all the application lists.

Since there may be multiple applications specified to recommend a pageIf a plurality of application ordered lists correspondingly exist, the invention can fuse the plurality of application ordered lists and remove the APP after the ordering to obtain the product containing N₂A second set of application alternatives for good quality APPs.

Optionally, the sub-step B11 includes:

substep B1101 selects the top P applications from each application ranking list.

Wherein the number of first P accumulations of the plurality of application ordered lists is greater than the number N of second application alternative set requirements₂And (4) respectively. For example, the application ordered list has 5 APPs, and the number of APPs is 5 × P.

Substep B1102, for each application, extracting a feature value of at least one dimension from a user behavior record associated with said application; the characteristic values of the at least one dimension include: the characteristic value of a first value dimension and the characteristic value of a user access behavior dimension;

and a substep B1103 of calculating a total eigenvalue of the application according to the eigenvalue of each dimension and the weight of the corresponding dimension.

In the embodiment of the present invention, the performing processes of sub-step B1102 and sub-step B1103 are similar to the performing processes of sub-step a10 and sub-step a11 mentioned in the first embodiment, and are not described in detail herein.

Through the sub-steps B1102 and B1103, the total characteristic value of each APP can be calculated.

Substep B1104, sorting the total eigenvalues by N₂The applications join a second set of application alternatives.

Then selecting N with the total characteristic values ranked at the top from each APP₂The applications join a second set of application alternatives.

Optionally, in the embodiment of the present invention, the fusion method may use a BORDA COUNT (nod COUNT) method to perform fusion, and the sub-step B11 may include:

and a sub-step B1111 of scoring the applications according to the positions of the applications in each application ranking list.

Assuming that the length of the result list of the crawling is N, its score in the list is denoted as N for the APP ranked first; for the second ranked APP, its score in the list is N-1; … … the APP ranked last has a score of 1. The above scoring process is repeated for all the APPs in the list.

Substep B1112, accumulating the scores of the applications in each application ranking list for the same application to obtain a total score of the applications;

the scores of the APPs in the lists are accumulated.

Substep B1113, ordering the total score by N₂The applications join a second set of application alternatives.

Sequencing the APP according to the accumulated scores from high to low, and then sequencing the total score to the top N₂The applications join a second set of application alternatives. Wherein N is₂The size of the integer is an integer greater than or equal to 0, and the size can be set according to actual requirements.

Of course, in the embodiment of the present invention, other fusion algorithms may also be adopted, and the embodiment of the present invention does not limit the present invention.

And step 230, fusing the first application candidate set and the second application candidate set to obtain a recommended application set.

After the two application alternative sets are obtained, the method needs to select a proper APP from the two application alternative sets and put the APP into the recommended application set, and the recommended application set is put into an application recommendation page to be displayed. Then, the present invention can merge the first application candidate set and the second application candidate set, and add the optimal multiple APPs to the recommended application set.

Optionally, the substep 230 comprises:

substep C11, top M of the set of applications is to be recommended₁Selecting the top M from the second application candidate set₁Each application is filled.

Substep C12, the remaining M of the set of applications will be recommended₂Position, selecting the top-ranked corresponding M from the first application candidate set₂Filling by the application of the user; wherein M is₁And M₂The sum is the number of applications in the set of recommended applications.

Before M for recommended application set₁Position, by the first M in the second application alternative set₁Filling APP; set of recommended applications M₁M after position₂And (4) sequencing the positions by the APP in the first application candidate set according to the total eigenvalue, and filling from high to low. This process is until the location of the set of recommended applications is filled. Wherein M is₁、M₂The value of (A) can be set according to actual needs, is usually 3 or 4, and generally does not exceed 10.

And 240, displaying each application in the recommended application set.

According to the embodiment of the invention, the quality of the APP is comprehensively evaluated through large-scale analysis of the user behavior records aiming at the APP display page, so that a first application alternative set is constructed. And then, acquiring the applications in each application recommendation page from a specified application recommendation page, such as an application recommendation page of an existing application management platform in a network, analyzing each application, constructing a second application alternative set, fusing the first application alternative set and the second application alternative set to obtain a final recommended application set, and then displaying the applications in the recommended application set in the application recommendation page. The embodiment of the invention integrates the access behaviors of the user, such as behaviors of browsing and downloading applications in different types of pages and application recommendation data in the existing specified application recommendation page, obtains a recommendation list of APP with higher quality through relatively objective statistics, reduces the labor cost and improves the recommendation efficiency.

EXAMPLE III

Referring to fig. 3, a flowchart illustrating steps of an embodiment of an application recommendation method according to the present invention is shown, which may specifically include the following steps:

step 310, obtaining user behavior records related to the applications and analyzing the applications to obtain a first application candidate set.

And 320, acquiring the applications in the application recommendation page of at least one application management platform, and analyzing each application to acquire a second application alternative set.

And 330, fusing the first application candidate set and the second application candidate set to obtain a recommended application set.

Step 340, for each application in the recommended application set, labeling the application with a classification label according to the classification label model.

Optionally, step 330 may be followed by:

and 300, analyzing the use behaviors of the functions of the application according to the users, and constructing an application classification label model.

Optionally, the method may further include:

step 340 and step 300.

In the embodiment of the invention, an APP classification label model is constructed in advance, and the APP classification label model is constructed by analyzing the use behaviors of each function of the application by each user.

In practical application, each APP usually has a plurality of segment function modules, and usually displays in the form of options and the like, for example, the APP of a game class also has corresponding function modules such as community forums, strategy communication and the like besides a game login entrance.

Then, for each APP, the embodiment of the present invention detects the usage behavior of the user on each function of the APP in each terminal that has downloaded the APP, and then constructs an APP classification tag model according to the usage behavior.

For example, if it is detected that a large number of users use a certain function of the APP, it may be considered that the users prefer to use the function of the APP, and a classification tag corresponding to the function may be marked for the APP. For example, if it is detected that APP "fast teeth" is downloaded by a large number of users and more than 70% of the users use the interactive game module therein, a category label "game you welcome" can be fitted according to the user behavior, and the category label is corresponding to "fast teeth".

In the embodiment of the invention, for classification labels such as 'white-collar-one', 'military fan', 'financing driver', 'knapsack client' and 'no classification', the classification labels can be regarded as a class, and then an APP classification label model, namely a maximum entropy classification label model, is constructed through a maximum entropy classifier in machine learning. The basic principle of the maximum entropy classifier is to seek a probability distribution which satisfies the maximum entropy of the existing sample set under a defined condition, and the construction process of the model is roughly as follows:

1. a series of classifications is preset. For example, the classification of "white-collar-group", "military fan", "financing carrier", "knapsack guest", "no classification" and the like, each classification is a classification label.

2. And selecting a series of APPs, and collecting the use behaviors of each terminal to each functional module of each APP.

3. And analyzing the use behaviors of the functional modules, determining the classification label corresponding to each APP, and taking the APP with the classification label as a training set.

4. Extracting characteristics of each APP in the training set, such as APP name, technical classification to which the APP belongs, keywords in APP description information, APP author, APP screenshot and the like. The technology to which the APP belongs is classified into a tool class, a system tool class, an input method class, and the like.

5. And inputting the characteristics of the APP into the maximum entropy classification label model as input values, and calculating the maximum entropy classification label model to obtain a first classification result.

6. And taking the classification label of each APP in the training set as a sample value, comparing the first classification result with the sample value, and if the error is greater than a threshold value, adjusting the model parameter of the maximum entropy classification label model by combining a training algorithm. And circularly training until the error between the first classification result and the sample value is less than the threshold value, and finishing constructing the maximum entropy classification label model.

Then, in step 340, for each application in the set of recommended applications, categorizing the application according to the categorical label model comprises:

and extracting the characteristics of each APP in the application recommendation set. Such as the aforementioned APP name, the technology class to which the APP belongs, keywords in the APP description information, the author of the APP, screenshots of the APP, and so on.

And then inputting the characteristics of each APP into the maximum entropy classification label model to obtain a classification result, and marking a classification label on the APP according to the classification result.

And if the classification result corresponds to a specific classification label, marking the classification label on the APP. And if the classification result corresponds to the classification result of 'no classification', not marking a classification label for the APP.

Of course, the APP classification label model in the embodiment of the present invention may also be constructed by a naive bayes classifier, a support vector machine classifier, a deep neural network, a random forest, and the like, which is not limited by the present invention.

Of course, in the embodiment of the present invention, the usage behavior of each APP on each function of the APP by the user in the user terminal may also be directly collected and analyzed, so as to determine the classification label of each APP. For example, for APP "fast teeth" which is downloaded by users in a large amount in the recent period of time, although it is software of a file transfer class, if 70% of users use the interactive game module therein, a category label of "game is popular you" is fitted according to the user behavior, and "fast teeth" is attached to the category label.

In the embodiment of the present invention, after obtaining the recommended application set, in order to enable the user to know APPs more quickly and to provide the user with a reason for recommendation, some classification labels may be applied to some APPs during actual display, for example: "white-collar clan", "military fan", "financing carrier", "spammer", and so on. The classification labels basically represent user groups suitable for the APP, and if the current user is just a certain person, the click rate of the user on the APP can be greatly improved. In the embodiment of the invention, for each application in the recommended application set, a classification label is marked on the application.

And 350, displaying each application in the recommended application set.

Through the steps, when the APP is displayed, the classification labels can be displayed outside basic information such as summary and download links of the APP, the user can know the APP quickly, and the content recommended to the user is enriched.

According to the embodiment of the invention, the quality of the APP can be comprehensively evaluated through large-scale analysis of the user behavior records aiming at the APP display page, so that the first application alternative set is constructed. Then, the applications in the application recommendation pages of the application presentation pages can be obtained from the application recommendation pages of the designated application presentation pages, such as the application presentation page application recommendation pages of the existing application management platform in the network, and the applications are analyzed to construct a second application alternative set, and then the first application alternative set and the second application alternative set are fused to obtain a final recommendation application set, and tag classification tags are printed for the applications in the recommendation application set, and then the applications in the recommendation application set are presented in the application presentation page application recommendation pages. The embodiment of the invention integrates the access behaviors of the user, such as behaviors of browsing and downloading applications in different types of pages and application recommendation data in the application recommendation page of the existing specified application display page, obtains a recommendation list of APP with higher quality by relatively objective statistics, reduces the labor cost, improves the recommendation efficiency, facilitates the user to quickly know the emphasized function of each application by the label classification label, and enriches the displayed content.

It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments of the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.

Example four

Referring to fig. 4, a block diagram of an embodiment of an application recommendation system according to the present invention is shown, and specifically includes the following modules:

the first application candidate set generating module 410 is configured to obtain a user behavior record related to an application and analyze each application to obtain a first application candidate set.

And a recommended application set obtaining module 420, configured to obtain a recommended application set according to the first application candidate set.

A presentation module 430, configured to present each application in the recommended application set.

Optionally, the first application candidate set generating module 410 includes:

the first initial characteristic word calculation module is used for extracting a characteristic value of at least one dimension from a user behavior record relevant to each application; the characteristic values of the at least one dimension include: the characteristic value of the first value dimension and the characteristic value of the user access behavior dimension.

And the first total characteristic value calculating module is used for calculating the total characteristic value of the application according to the characteristic value of each dimension and the weight of the corresponding dimension.

A first alternative set construction module for ordering the total eigenvalues by N₁Each application joins the first set of application alternatives.

Optionally, the first initial feature word calculating module includes:

and the first value dimension characteristic value calculation module is used for calculating the proportion between the first value data of the application and the distribution amount of the application for each application, and taking the proportion as the characteristic value of the first value dimension.

Optionally, the dimension characteristic value of the user access behavior includes: browsing the characteristic value of the dimension, and downloading the characteristic value of the dimension.

Optionally, when the feature value of the user access behavior dimension is a feature value of a browsing dimension, the first initial feature word calculation module includes:

and counting, for each application, a first browsing number of times that the application is browsed in the application summary display page, a second browsing number of times that the application is browsed in the search result page, and a third browsing number of times that the application is browsed in the application detail display page.

And the browsing characteristic value calculating module is used for respectively combining the first downloading times, the second downloading times and the third downloading times with the weight of the page where the first downloading times, the second downloading times and the third downloading times are located, and calculating the characteristic value of the downloading dimension.

Optionally, when the feature value of the user access behavior dimension is a feature value of a download dimension, the first initial feature word calculation module includes:

and the page downloading frequency counting module is used for counting the first downloading frequency of the application summary display page, the second downloading frequency of the application summary display page and the third downloading frequency of the application detail display page.

And the download characteristic value calculating module is used for respectively combining the first download times, the second download times and the third download times with the weight of the page where the download dimensions exist and calculating the characteristic value of the download dimensions.

EXAMPLE five

Referring to fig. 5, a block diagram of an embodiment of an application recommendation system according to the present invention is shown, and specifically includes the following modules:

the first application candidate set generating module 510 is configured to obtain a user behavior record related to an application and analyze each application to obtain a first application candidate set.

A second application alternative set generating module 520, configured to obtain applications in an application recommendation page of at least one application management platform and analyze each application to obtain a second application alternative set;

the recommended application set obtaining module 530 specifically includes:

and a fusion module 531, configured to fuse the first application candidate set and the second application candidate set to obtain a recommended application set.

A presentation module 540, configured to present each application in the recommended application set.

Optionally, the second application candidate set generating module 520 includes:

and the ordered list acquisition module is used for acquiring a corresponding application ordered list for the application recommendation page of at least one application management platform.

A list fusion module used for performing fusion calculation on each application ordered list to obtain a list N when the application ordered lists are two or more₂A second set of application alternatives for each application, wherein N₂Is less than or equal to the sum of the number of applications in all the application lists.

Optionally, the list fusion module includes:

the application extraction module is used for selecting P applications which are ranked at the front from each application ranking list;

the second initial characteristic word calculation module is used for extracting a characteristic value of at least one dimension from the user behavior record relevant to the application aiming at each application; the characteristic values of the at least one dimension include: the characteristic value of the first value dimension and the characteristic value of the user access behavior dimension.

And the second total characteristic value calculating module is used for calculating the total characteristic value of the application according to the characteristic value of each dimension and the weight of the corresponding dimension.

A second alternative set construction module for ordering the total eigenvalues by N₂The applications join a second set of application alternatives.

Optionally, the list fusion module includes:

and the first scoring module is used for scoring each application according to the position of the application in each application ranking list.

And the scoring accumulation module is used for accumulating the scores of the applications in each application ranking list aiming at the same application to obtain the total score of the applications.

A ranking construction module for ranking the total score by N₂The applications join a second set of application alternatives.

Optionally, the fusion module 531 includes:

a first filling module for recommending the top M of the application set₁Position, selecting ranking from the second application alternative setFront M₁Each application is filled.

A second filling module for recommending the rest M of the application set₂Position, selecting the top-ranked corresponding M from the first application candidate set₂Is filled by an application of M₁And M₂The sum is the number of applications in the set of recommended applications.

EXAMPLE six

Referring to fig. 6, a block diagram of an embodiment of an application recommendation system according to the present invention is shown, and specifically includes the following modules:

the first application candidate set generating module 610 is configured to obtain a user behavior record related to an application and analyze each application to obtain a first application candidate set.

And a second application candidate set generating module 620, configured to obtain applications in the application recommendation page of the at least one application management platform and analyze the applications to obtain a second application candidate set.

The recommended application set obtaining module 630 specifically includes:

the fusion module 631 is configured to fuse the first application candidate set and the second application candidate set to obtain a recommended application set.

A classification module 640, configured to, for each application in the recommended application set, apply a classification label to the application according to the classification label model.

A presentation module 650, configured to present each application in the recommended application set.

Optionally, the method further includes:

and the classification label model building module is used for analyzing the use behaviors of the functions of the application according to the users and building an application classification label model.

For the system embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.

The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, embodiments of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

In a typical configuration, the computer device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory. The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or FLASH memory (FLASH RAM). Memory is an example of a computer-readable medium. Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, computer readable MEDIA does not include non-TRANSITORY computer readable MEDIA (transport MEDIA), such as modulated data signals and carrier waves.

Embodiments of the present invention are described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal, create a system for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction system which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing terminal to cause a series of operational steps to be performed on the computer or other programmable terminal to produce a computer implemented process such that the instructions which execute on the computer or other programmable terminal provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

While preferred embodiments of the present invention have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including preferred embodiments and all such alterations and modifications as fall within the scope of the embodiments of the invention.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.

The application recommendation method and the application recommendation system provided by the invention are described in detail, and the principle and the implementation mode of the invention are explained by applying specific examples, and the description of the examples is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. An application recommendation method, comprising:

displaying each application in the recommended application set;

before the step of obtaining the recommended application set according to the first application candidate set, the method further includes:

the step of obtaining a recommended application set according to the first application candidate set includes:

fusing the first application alternative set and the second application alternative set to obtain a recommended application set;

the step of fusing the first application candidate set and the second application candidate set to obtain the recommended application set includes:

will recommend the rest of the application set₂Selecting the corresponding M in the first application candidate set₂Filling by the application of the user; wherein M is₁And M₂The sum is the number of applications in the set of recommended applications.

2. The method of claim 1, wherein the step of obtaining a record of user behavior associated with the application and analyzing the applications to obtain a first set of application alternatives comprises:

3. The method according to claim 2, wherein the step of extracting, for each application, a feature value of at least one dimension from a user behavior record associated with the application comprises:

4. The method of claim 2, the user accessing feature values of a behavior dimension comprising: browsing the characteristic value of the dimension, and downloading the characteristic value of the dimension.

5. The method of claim 4,

when the feature value of the user access behavior dimension is a feature value of a browsing dimension, the step of extracting, for each application, a feature value of at least one dimension from a user behavior record associated with the application includes:

6. The method of claim 4, wherein when the feature value of the user access behavior dimension is a feature value of a download dimension, the step of extracting, for each application, a feature value of at least one dimension from a user behavior record associated with the application comprises:

7. The method according to claim 1, wherein the step of obtaining the applications in the application recommendation page of at least one application management platform and analyzing the applications to obtain the second application alternative set comprises:

8. The method according to claim 7, wherein when the application ordered list is two or more, the fusion calculation is performed on each application ordered list to obtain a fusion calculation including N₂The step of the second set of application alternatives for each application comprises:

selecting P applications with the top rank from each application ranking list;

9. The method of claim 1, after obtaining the set of recommended applications, further comprising:

10. An application recommendation system, comprising:

the display module is used for displaying each application in the recommended application set;

wherein, still include:

the second application alternative set generation module is used for acquiring the applications in the application recommendation page of at least one application management platform and analyzing each application to acquire a second application alternative set;

the recommended application set obtaining module specifically includes:

the fusion module is used for fusing the first application alternative set and the second application alternative set to obtain a recommended application set;

wherein the fusion module comprises:

a first filling module for recommending the top M of the application set₁Selecting the top M from the second application candidate set₁Filling by each application;

a second filling module for recommending the rest M of the application set₂Selecting the corresponding M in the first application candidate set₂Is filled by an application of M₁And M₂The sum is the number of applications in the set of recommended applications.