CN102567511A - Method and device for automatic application recommendation - Google Patents

Abstract

The application provides a method and a device for automatic application recommendation, wherein the method comprises the following steps: receiving an application acquisition request submitted by a user from a client, wherein the application acquisition request comprises a user identifier; extracting the existing user behavior information of the corresponding user from a user feature library according to the user identification, wherein the user behavior information comprises operation information of the user aiming at the previously recommended application; determining the application type recommended to the user according to the user behavior information; extracting matched applications in the application data set of the application category according to operation information of a user aiming at the previously recommended applications; and generating a corresponding application folder according to the application category, and putting the matched application into the corresponding application folder for recommendation. The method and the device can meet the individual requirements of the user, and improve the recommendation efficiency and the coverage rate.

Description

Method and device for automatic application recommendation

technical field

The present application relates to the technical field of information processing, in particular to a method for automatic application recommendation and an apparatus for automatic application recommendation.

Background technique

The Internet is an important way for people to obtain information. The main feature of the traditional Internet is that when users look for things they are interested in, they need to conduct a large number of searches through the browser, and at the same time need to manually filter out a large number of irrelevant results. The operation is cumbersome and Takes time and energy.

With the rapid development of Internet technology, people have more and more demands on various network applications. The deployment of applications on the client side is becoming more and more bloated, which not only causes a waste of terminal resources, but also makes it difficult to manage. Even if the client-server architecture is used for deployment management, the server side also lacks the ability to manage subsequent use after the deployment of the client is completed.

Although the concept of the so-called "thin client (Thin Client)" has appeared now, the thin client transmits its mouse, keyboard and other inputs to the server for processing, and the server returns the processing results to the client for display. However, this processing mode is limited by the network transmission speed and the processing capacity of the server. Therefore, it is more used in enterprise-level commercial LANs, and it is not suitable for the entertainment needs of ordinary users.

In order to provide users with a better user experience, the prior art proposes a scheme of automatically recommending applications of interest to users, that is, actively recommending and providing applications of interest to users by knowing their interests. However, this method of application recommendation is mainly manually recommended by editors. This method of manual recommendation by editors mainly has the following defects:

1. The efficiency is too low, and the recommendation coverage rate for applications is too low. For example, for hundreds of thousands of applications on the platform, manual recommendation can only recommend a few hundred every day. It is actually impossible to recommend all applications, and the coverage is too low because the proportion is too low.

2. This kind of recommendation is completely based on the unified recommendation principle of editors, which is the same for every user and cannot meet the individual needs of users. Because some recommended applications are suitable for some users, but not for some users.

Therefore, a technical problem that needs to be urgently solved by those skilled in the art is to propose a mechanism for automatic application recommendation to meet the individual needs of users and improve recommendation efficiency and coverage.

Contents of the invention

The technical problem to be solved in this application is to provide a method for automatic application recommendation to meet the individual needs of users and improve recommendation efficiency and coverage.

The present application also provides a device for automatic application recommendation to ensure the application and realization of the above method in practice.

In order to solve the above problems, the embodiment of the present application discloses a method for automatic application recommendation, including:

receiving an application acquisition request submitted by a user from a client, where the application acquisition request includes a user identifier;

Extracting existing user behavior information of the corresponding user from the user feature database according to the user identifier, the user behavior information includes the user's operation information for the previously recommended application;

Determine the application category recommended to the user according to the user behavior information;

In the application data set of the application category, extract matching applications according to the user's operation information on previously recommended applications;

A corresponding application folder is generated according to the application category, and the matching application is put into the corresponding application folder for recommendation.

Preferably, the method may also include:

Collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID.

Preferably, the user behavior information also includes the user's local operation behavior information, and/or, the user's online operation behavior information;

The step of determining the application category recommended to the user according to the user behavior information may include:

Extracting classification labels and corresponding first operation frequencies from the user's local operation behavior information and/or online operation behavior information;

Converting the classification label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of the classification label and the application category;

Extracting application information and a corresponding second operation frequency operated by the user within a preset time period from the user's operation information on previously recommended applications, the application information including application categories;

Calculate the weight of each application category according to the first operation frequency and the second operation frequency, and sort according to the weight of the application category from high to low;

A preset number of first n application categories are extracted as application categories recommended to the user; wherein, n is a positive integer greater than 1.

Preferably, the application data set of a certain application category can be generated through the following steps:

Obtain applications of the same application category, where the applications have classification labels;

Determine the main application and the application to be recommended in the application, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application;

Obtaining quality scoring parameters of the application to be recommended;

Extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the first m applications to be recommended by a preset number; wherein, m is A positive integer greater than 1;

The main application and the extracted corresponding applications to be recommended are combined into an application data set of the current application category.

Preferably, in the application data set of the application category, the step of extracting the matching application according to the user's operation information on the previously recommended application may include:

According to the user's operation information on the previously recommended application, count the main application and the corresponding third operation frequency, the main application is the application operated by the user;

In the application data set corresponding to the application category, the matching applications to be recommended are extracted according to the main application, and among the matched applications to be recommended, a certain number of applications to be recommended are respectively extracted using the third operation frequency as the weight of application extraction. For recommending applications, a total of applications to be recommended that satisfy the first preset number are extracted.

Preferably, in the application data set of the application category, the step of extracting the matching application according to the user's operation information on the previously recommended application may further include:

Obtain the application category corresponding to the main application, sort the main application according to the third operation frequency in the same application category, and extract the first k main applications of the preset number; wherein, the k is a positive number greater than 1 integer;

Pairing the extracted main applications in pairs, calculating the total number of simultaneous occurrences of the paired main applications, and generating frequent 2-itemsets;

Calculate the number of occurrences of each main application alone to generate frequent 1-itemsets;

Calculate the confidence of each main application according to the frequent 2-itemset and frequent 1-itemset, and sort the main applications according to the confidence;

The extracted applications to be recommended satisfying the first preset quantity are matched with the main applications sorted by confidence to generate a final recommended matching application.

The embodiment of the present application also discloses a device for automatic application recommendation, including:

A request receiving module, configured to receive an application acquisition request submitted by a user from a client, where the application acquisition request includes a user identifier;

The previous behavior information extraction module is used to extract the existing user behavior information of the corresponding user from the user feature database according to the user identification, and the user behavior information includes the user's operation information for the previous recommended application;

An application category determination module, configured to determine the application category recommended to the user according to the user behavior information;

A matching application acquisition module, configured to extract matching applications according to the user's operation information on previously recommended applications in the application data set of the application category;

The application recommendation module is configured to generate a corresponding application folder according to the application category, and put the matching application into the corresponding application folder for recommendation.

Preferably, said device also includes:

The behavior statistics module is used to collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID.

Preferably, the user behavior information also includes the user's local operation behavior information, and/or, the user's online operation behavior information;

The application category determination module may include:

The first feature extraction submodule is used to extract classification labels and corresponding first operation frequency from the user's local operation behavior information and/or online operation behavior information;

A conversion submodule, configured to convert the classification label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of a classification label and an application category;

The second feature extraction sub-module is used to extract the application information and the corresponding second operation frequency operated by the user within the preset time period from the operation information of the user on the previously recommended application, the application information includes application category;

A sorting submodule, configured to calculate the weight of each application category according to the first operation frequency and the second operation frequency, and sort according to the weight of the application category from high to low;

The category selection sub-module is used to extract a preset number of top n application categories as application categories recommended to users; wherein, n is a positive integer greater than 1.

Preferably, said device may also include:

The application data set generation module is used to generate application data sets of various application categories: specifically including:

The similar application acquisition sub-module is used to acquire applications of the same application category, and the applications have classification tags;

The similarity calculation sub-module is used to determine the main application and the application to be recommended in the application, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application;

A quality scoring parameter acquisition submodule, configured to acquire the quality scoring parameters of the application to be recommended;

The application to be recommended extraction sub-module is used to extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the top m to be recommended Recommended applications; wherein, the m is a positive integer greater than 1;

The application data set forming sub-module is used to form the main application and the corresponding extracted application to be recommended into an application data set of the current application category.

Preferably, the matching application acquisition module may include:

The main application statistics sub-module is used to count the main application and the corresponding third operation frequency according to the user's operation information on the previously recommended application, and the main application is an application operated by the user;

The application-to-be-recommended determination submodule is configured to extract a matching application to be recommended according to the main application in the application data set corresponding to the application category, and use the third operation frequency as the application in the matched application to be recommended A certain number of applications to be recommended are respectively extracted by the extracted weights, and applications to be recommended that satisfy the first preset number are extracted in total.

Preferably, the matching application acquisition module may also include:

The main application selection sub-module is used to obtain the application category corresponding to the main application, sort the main applications according to the third operation frequency within the same application category, and extract the first k main applications of the preset number; wherein, The k is a positive integer greater than 1;

The frequent 2-itemset calculation submodule is used to pair the extracted main applications in pairs, calculate the total number of simultaneous occurrences of the paired main applications, and generate frequent 2-itemsets;

The frequent 1-itemset calculation sub-module is used to calculate the number of occurrences of each main application and generate frequent 1-itemsets;

Confidence degree calculation sub-module, used to calculate the confidence degree of each main application according to the frequent 2-itemset and frequent 1-itemset, and sort the main applications according to the confidence degree;

The matching application determining submodule is configured to match the extracted applications to be recommended that meet the first preset number and the main applications sorted by confidence to generate a final recommended matching application.

Compared with the prior art, the present application has the following advantages:

Based on the applications that have been recommended to the user, this application analyzes the user's operation information on the previously recommended application, combines the user's online operation behavior information and/or local operation behavior information, and determines the application category preferred by the user behavior information, and then In the application data set corresponding to the application category, based on the user's operation information on the previously recommended application, combined with the user's online operation behavior information and/or local operation behavior information, the applications that best meet the user's interests are extracted, and these applications are placed in the It can be recommended in the folder corresponding to the application category, so as to establish a connection between the application and the user, fully meet the individual needs of the user, and effectively improve the recommendation efficiency and coverage of the application.

Furthermore, this application uses the user interface as an entry, and recommends applications to the user directly on the interface or through the links on the interface through the application folder icon, so that the user can obtain the desired application faster and easier, and facilitates the user's operation; and , the user can be prompted to use the application by using the icon as the application entry, but the configuration file corresponding to the application is not actually installed before the user actually chooses to use it. In this way, it does not occupy too much client resources before use. In addition, the icons in the user interface can be centrally deployed or pushed by the central server on the network side, which prevents malicious programs from randomly adding malicious icons in the interface, further improving security.

Description of drawings

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

Fig. 2 is a structural block diagram of an embodiment of an application automatic recommendation device of the present application.

Detailed ways

In order to make the above objects, features and advantages of the present application more obvious and comprehensible, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

The core idea of the embodiment of the present application is to analyze the user's operation information on the previously recommended application based on the applications that have been recommended to the user, and determine the user behavior information in combination with the user's online operation behavior information and/or local operation behavior information The preferred application category, and then in the application data set corresponding to the application category, according to the user's operation information on the previously recommended application, combined with the user's online operation behavior information and/or local operation behavior information, extract the most suitable for the user's interests The applications are put into folders of corresponding application categories for recommendation, so as to establish a connection between the application and the user.

Referring to FIG. 1 , it shows a flow chart of the steps of an embodiment of an application automatic recommendation method of the present application, which may specifically include the following steps:

Step 101. Receive an application acquisition request submitted by a user from a client, where the application acquisition request includes a user ID;

In a specific implementation, the user starts the client to trigger the application acquisition request, and the user can also manually trigger the application acquisition request, which is not limited in this application.

Step 102, extracting existing user behavior information of the corresponding user from the user feature database according to the user identifier, the user behavior information including the user's operation information for the previously recommended application;

The following information may be recorded in the user feature database: user identifier Mid, classification label tag of user behavior information, and corresponding operation frequency weight.

In a preferred embodiment of the present application, the user's behavior information may include the user's local operation behavior information, and/or the user's online operation behavior information, and the user's operation information on previously recommended applications. The user's local operation behavior information and online operation behavior information usually have classification label (tag) information, for example, for the video opened by the user's local operation, it contains Naruto, animation, series, fantasy, adventure, Kishimoto, etc. Classification label information such as Qi History; or, for example, for websites visited by users on the Internet, classification label information such as videos, movies, comedy movies, and king of comedy. The application also has application category and classification label information.

The user's local operation behavior information and online operation behavior information can be collected by client software installed on the user equipment, wherein the user equipment can include various types of smart devices such as computers, notebook computers, mobile phones, PDAs, and tablet computers. terminal. The following provides several examples of collecting user’s local operation behavior information, and/or, user’s online operation behavior information:

Example 1, collect the user's online operation behavior information for a period of time through the browser, including the URLs visited and the corresponding number of visits, etc.;

For example, the online operation behavior information collected by the user within 15 days through the browser is:

Example 2, the user’s local operation behavior information is collected through the security software installed on the user’s device. For example, the user’s online operation behavior information and local behavior information within 15 days are collected through 360 Network Shield: open Baofengyingyin and its times, open a certain Games and their times etc.

Of course, the above collection method and collected information are only used as examples, and it is feasible for those skilled in the art to use any method to collect required user behavior information according to the actual situation, and this embodiment of the present application does not need to limit it.

Step 103, determining the application category recommended to the user according to the user behavior information;

In a preferred embodiment of the present application, the step 103 may specifically include the following sub-steps:

Sub-step S11, extracting classification labels and corresponding first operation frequency from the user's local operation behavior information and/or online operation behavior information;

Sub-step S12, converting the category label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of a category label and an application category;

Sub-step S13, extracting the application information operated by the user within a preset time period and the corresponding second operation frequency from the user's operation information on the previously recommended application, the application information including the application category;

Sub-step S14, calculating the weight of each application category according to the first operation frequency and the second operation frequency, and sorting the weights of the application categories from high to low;

Sub-step S15 , extracting a preset number of top n application categories as application categories recommended to the user; wherein, n is a positive integer greater than 1.

In practice, the application category that the user behavior information conforms to can be obtained by analyzing the user behavior information according to the application category preset by the technician. For example, there are 20 basic categories of pre-set application folders, and by analyzing user behavior information, it is found that some basic categories are not needed for the current user, and the application category to which the user behavior information belongs can be classified as closer to the user before. 3 or 5 of behavioral habits. For example, video, games, education, etc.

In order to enable those skilled in the art to better understand the present application, the following uses a specific example to illustrate the process of determining the application category recommended to the user based on user behavior information:

Data Sources:

(1) NetShield data in the last 15 days: Data11;

(2) In the last 15 days, the user added or clicked on the application of the secure desktop: Data12;

The data format of Data11 is:

User ID Mid Category Label Interest First Operation Frequency weight1

The data example of Data11 is as follows:

0000175873530b93d848614a0c188c5b novel-dm 1;

000020218613d5fc8e05c314dba32956 comic-dm 4;

00002e3bb9037870973b328078971c98 4399-dm 1;

The data format of Data12 is:

original log

The data example of Data12 is as follows:

123.97.168.210--[15/Oct/2011:23:00:01+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝102000032&fenleiid＝4&from＝0&leixing＝2&style＝fullscreen&uid＝1&pid＝softmgr&m＝45c06dc58f5ccb64c162b646fcecc541&modulever =1.4.0.1103&appver=1.4.0.1103HTTP/1.1″2000″-″″Mozilla/4.0 (compatible; MSIE 7.0; Windows NT 6.1; Trident/4.0; GTB7.1; SLCC2; .NET CLR 2.0.50727; . NET CLR 3.5.30729; .NETCLR 3.0.30729; Media Center PC 6.0)″

Step1: Extract the classification label Interest from the network shield data Data11, and convert it into the basic classification of user interests under the application folder classification system through the preset conversion rule table (yunCatToZhuoMianCat.conf), that is, convert the category label into the corresponding application category. The format of the preset conversion rule table yunCatToZhuoMianCat.conf may include: classification label, application category name AppName and application category identifier Appid information, as shown in the following table:

The results of Data11 conversion are shown in the table below:

Mid Interest AppName Appid Weight1 000020218613d5fc8e05c314dba32956 comic-dm fashion entertainment 8 4 0000175873530b93d848614a0c188c5b novel-dm novel 11 1 00002e3bb9037870973b328078971c98 4399-dm game 5 1

Step2: By analyzing the original log Data12, calculate the operation frequency (second operation frequency) of clicking or adding each application in the last 15 days for each Mid, and determine the application category corresponding to the user's interest according to the Appid_name classification comparison table.

Among them, the format of the Appid_name classification comparison table is as follows:

Application identification Appid application category identification fenleiid classification label tag application name AppName.

The data example of the Appid_name classification comparison table is as follows:

100026002 5 Small Games Other Chess Card Cartoon English Single Tractor

100013330 4 TV dramas Ancient costumes Plot Scheming

100114314 6 Health Family Doctor Self-diagnosis Family Doctor

100114370 6 gourmet braised pork recipe cookbook recipe cooking braised pork

100013349 4 TV drama military suspense informer

By analyzing the original log Data12, calculate the operation frequency (second operation frequency Weight2) of clicking or adding each application for each Mid in the last 15 days, as shown in the following table:

Refer to the Appid_name category comparison table above to determine the application category corresponding to the user's interest, as shown in the following table:

Mid fenleiid Weight2 00008fc5c27c3354e1e0c9b6b7527dd9 5 1 0000b5d11c0c8ea46817fc32f467c3ba 4 3 0001555e4ea2b299b6fbc55f46eeb771 6 4

Step3: The results of Step1 and Step2 are weighted and averaged according to the first operation frequency and the second operation frequency, and then sorted according to the final score. Take top9 as the application category recommended to the user, that is, the application category of the classified application folder displayed finally .

For example: for a certain Mid, the result of Step1 is: type1 clicks n1 times, type2 clicks n2 times, type3 clicks n3 times...;

The result of step2 is: type1 acts N1 times, type2 clicks N2 times, type3 clicks N3 times...then score1=n1*0.6+N1*0.4, score2=n2*0.6+N2*0.4, score3=n3*0.6 +N3*0.4...

Sort by score, and take the top 9 application categories as the application categories recommended to the current user.

In the specific implementation, if the application categories divided by the analysis of user behavior information cannot reach the specified number, if only three categories can be generated using the above example, which cannot meet the requirements of nine application categories, then the network can be calculated according to the cloud statistics. The application category actually used by the user the most or the latest application category is used as the recommended application category for completion.

Of course, the above method of classifying the categories of user behavior information is only used as an example, and it is feasible for those skilled in the art to adopt a method according to the actual situation. Convert to application categories according to preset rules; or directly extract classification tags as application categories, etc., which is not limited in this application.

Step 104, in the application data set of the application category, extract matching applications according to the user's operation information on previously recommended applications;

The application (Application) refers to various services used by users on the network, such as application programs, web pages, videos, novels, music, games, news, shopping, and mailboxes. The application data set contains multiple applications, which come from various open platforms. In the embodiment of this application, the application will carry category information (application category) and some classification labels.

In a preferred embodiment of the application, the application data set of a certain application category can be generated through the following sub-steps:

Sub-step S21, obtaining applications of the same application category, the applications have classification tags;

Sub-step S22. Determine the main application and the application to be recommended among the applications, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application;

Sub-step S23, acquiring the quality scoring parameters of the application to be recommended;

Sub-step S24, respectively extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the first m applications to be recommended by a preset number; among them, The m is a positive integer greater than 1;

Sub-step S25, compose the main application and the corresponding extracted application to be recommended into an application data set of the current application category.

In the above preferred embodiment, for applications of the same application category, the similarity between applications is calculated according to their classification labels to form an application data set including the main application and the application to be recommended.

In order for those skilled in the art to better understand the present application, a specific example is used below to illustrate the above-mentioned process of generating the application data set.

1). Calculate the similarity between applications according to the application category of the application and the classification label tag of the application:

The data format of the input data is: Appid fenleiID tag1 tag2 tag3 tag4tag5 tag6 tag7 tag8......;

The data format of the result file Data1 is: the main application ID (main Appid), the application ID to be recommended (Appid to be recommended), similarity

The calculation method of similarity is:

In the same application category, the previous i tags are used as class marks. In the same tag, apps are combined in pairs to calculate their similarity. The calculation formula is: Similarity=i/(n1+n2-i); among them, n1 is the number of tags behind app1 (application 1), and n2 is the number of tags behind app2 (application 2); the minimum value of i is 2, and the maximum value is n1. For example:

The input data is:

100030071 4 Movie Drama Comedy Love Wu Chenjun Liu Yanjun Xie Xiaoming Others 2010 Mainland China

100030073 4 Movie Drama Comedy Action Kim Ha Neul Kang Ji Hwan Shin Tae Ra Others 2009 South Korea

100030074 4 Movie Mystery Science Fiction Thriller Nicolas Cage Chandler Canterbury Alex Proyas Others 2009 USA

 …

The result file Data1 is:

100030071 100030073 0.25

100030071 100030074 0.11

100030073 100030074 0.11

 …

2). Within the same main Appid, comprehensively sort the recommended Appids according to their similarity and Appid's quality score (daily downloads, user ratings), that is, Appid similarity * similarity weight + Appid quality score * (1- similarity weight), intercept the top 50 Appids to be recommended with the highest comprehensive scores, and then merge them into one row;

Input data: Data1 (the result file of the previous step)

The format of the output data Data2 is as follows: main Appid to be recommended Appid1 to be recommended Appid2 to be recommended Appid3 to be recommended Appid4...

For example:

The input data Data1 is:

100030071 100030073 0.25

100030071 100030074 0.11

100030073 100030074 0.11

100030073 100030071 0.25

100030074 100030071 0.11

100030074 100030073 0.11

 …

The output data Data2 is:

100030071 100030073 100030074...

100030073 100030071 100030074...

100030074 100030071 100030073...

In a preferred embodiment of the application, the step 104 may further include the following sub-steps:

Sub-step S31, according to the user's operation information on the previously recommended application, count the main application and the corresponding third operation frequency, the main application is the application operated by the user;

Sub-step S32, in the application data set corresponding to the application category, extract the matching application to be recommended according to the main application, and extract the third operation frequency as the weight of application extraction from the matching application to be recommended For a certain number of applications to be recommended, a total of applications to be recommended that satisfy the first preset number are extracted.

In order for those skilled in the art to better understand the present application, the above sub-steps S31-S32 are described below through a specific example.

Step1:

3). According to Mid's application operation behavior log, count the number of times Mid added or clicked on each app;

Input data: Mid added or clicked on the log record of the application (the application operation behavior log in the last 30 days);

The format of the output data Data3 is: Mid Main Appid (the id of the clicked or added app)

weight3 (third operation frequency)

If the input data is:

27.185.166.230--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝103352&fenleiid＝10001&from＝0&leixing＝1&style＝fullscreen&uid＝1&pid＝h_home_inst&m＝71ddd8f9f1c84e16438ef109f4b6d77b&modulever =1.4.0.1041&appver=1.4.0.1041 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 7.0; Windows NT 6.0; SLCC1; .NET CLR2.0.50727; Media Center PC 5.0; .NET CLR 3.5. 30729;.NET CLR 3.0.30618)″

111.127.218.150--[20/Aug/2011:10:11:47+0800] "GET/stat.html?type=open&action=tianjiayingyong&Appid=100018815&fenleiid=4&sort=%b6%af%bb%ad&from=5&style=fullscreen&uid= 1&pid=h_home_inst&m=9e236bafe13c8348247781c2d0fab7a7&modulever=1.0.2.1025&appver=1.4.0.1040 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1; SV1; 4399) 9″ Box.

58.50.201.130--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝103352&fenleiid＝10001&from＝6&leixing＝1&style＝iphone&uid＝1&pid＝h_home&m＝3d3e77348ff2fbfa6af7c3751a00edae&modulever =1.4.0.1040&appver=1.4.0.1040HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1)″

110.178.40.7--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝100000525&fenleiid＝4&from＝0&leixing＝2&style＝fullscreen&uid＝1&pid＝softmgr&m＝f16b5a2c01d64fcfa3ff5f035ce74677&modulever =1.4.0.1041&appver=1.4.0.1041 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 7.0; Windows NT 5.1; Trident/4.0)″

58.50.201.130--[20/Aug/2011:10:11:47+0800] "GET/stat.html?type=open&action=zuixiaohuazhuomian&count=90468&uid=1&pid=h_home&m=3d3e77348ff2fbfa6af7c3751a00edae&modulever1=1.4.HTTP00. /1.1″200 0″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1)″

 …

The output data Data3 is:

0004a218f3b8a96b59f67a8f14be5e98 100000289 102

0004ce91dc7726afdc420a97ad050f5a 100103087 1

0004cfe7c72aa83bfa20e561b6a00823 102020214 2

0004ec5290131cf7fa6d5a18833e06a5 102005903 3

0004f55c24b9fb8745b6d7595d94972d 120033743 1

00051004579f67de2066dc94f8952fd4 100000275 5

00054c7ba3cf1b45330444cbb737cac4 100114758 4

 …

4). According to Data2 and Data3, match through the main app, and then sort according to weight

Mid1 fenleiid1 Appid1 Appid2 Appid3 Appid4...weight

Mid1 fenleiid2 Appid11 Appid22 Appid33 Appid44...weight

According to the weight, the greater the weight, the more apps are intercepted in this category. A total of 50 Appids are intercepted by random interception, and then merged with Mid and fenleiid as keywords, and the generated result file is

Mid1 fenleiid Appid1 Appid2 Appid3 Appid4...Appid20

For example:

Input data: Data2 and Data3

Among them, Data2 is:

001 (100 101 102 103 104 105 106 107 108 109...)

002(201 202 203 204 205 206 207 208 209 210...)

003 (301 302 303 304 305 306 307 308 309 310...)

008 (801 802 803 804 805 806 807 808 809 810...)

 …

Assume that the part in the brackets in the above Data2 is the top 50 Appids with the highest similarity with the main app

Data3 is:

Xx1 00 15

Xx1 002 3

Xx1 003 2

Xx1 008 5

 …

Through the Appid_name classification comparison table, its category is mapped out, and the obtained Data33 is:

Xx1 1 001 5

Xx1 1 002 3

Xx1 1 003 2

Xx1 2 008 5

 …

Intermediate output data:

The first data: Xx1 1 100 101 102 103 104 105 106 107 108 109...5

The second data: Xx1 1 201 202 203 204 205 206 207 208 209 210......3

The third data: Xx1 1 301 302 303 304 305 306 307 308 309 310...2

The fourth data: Xx1 2 801 802 803 804 805 806 807 808 809 810......5

 …

According to the weight calculation method:

5/(5+3+2)*20=10 Appids are randomly selected from the first data;

Randomly select 3/(5+3+2)*20＝6 Appids from the 2nd data;

Randomly select 2/(5+3+2)*20＝4 Appids from the third data;

The final output data is:

Xx1 1 100 101 102 103 104 105 106 107 108 109 204 205 206 207 208 209306 307 308 303

Xx2 2 801 802 803 804 805 806 807 808 809 810......

 …

More preferably, said step 104 may further include the following sub-steps:

Sub-step S33: Obtain the application category corresponding to the main application, sort the main applications according to the third operation frequency within the same application category, and extract the first k main applications of the preset number; wherein, k is A positive integer greater than 1;

Sub-step S34, pairing the extracted main applications in pairs, calculating the total number of simultaneous appearances of the pairwise main applications, and generating frequent 2-itemsets;

Sub-step S35, calculating the number of occurrences of each main application alone, and generating frequent 1-itemsets;

Sub-step S36, calculating the confidence of each main application according to the frequent 2-itemset and frequent 1-itemset, and sorting the main applications according to the confidence;

Sub-step S37 , matching the extracted applications to be recommended that meet the first preset number and the main applications sorted by confidence, to generate a final recommended matching application.

One of the core concepts of this embodiment is to find similar applications based on the user's behavior of adding or clicking applications, that is, his preference for applications, and then recommend similar applications to the user according to his historical preferences, that is, the applications that were added or clicked in history. From a calculation point of view, it is to use all users' preferences for a certain application as a vector to calculate the similarity between applications. After obtaining the similar applications of the application, predict the application that the current user has not expressed preference according to the user's historical preference. Compute to get a sorted list of apps as recommendations.

In order for those skilled in the art to better understand the present application, the above sub-steps S33-S37 are described below through a specific example.

Step2:

1). According to the behavior log of Mid adding and clicking on the application, count the number of times Mid added or clicked on each app;

Input data: Mid add or click on the application log (application operation behavior log in the last 30 days);

The format of the output data Data1 is: Mid main Appid (app clicked or added)

weight3 (third operation frequency)

For example: the input data is:

27.185.166.230--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝103352&fenleiid＝10001&from＝0&leixing＝1&style＝fullscreen&uid＝1&pid＝h_home_inst&m＝71ddd8f9f1c84e16438ef109f4b6d77b&modulever =1.4.0.1041&appver=1.4.0.1041 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 7.0; Windows NT 6.0; SLCC1; .NET CLR2.0.50727; Media Center PC 5.0; .NET CLR 3.5.30729 ;.NET CLR 3.0.30618)"

111.127.218.150--[20/Aug/2011:10:11:47+0800] "GET/stat.html?type=open&action=tianjiayingyong&Appid=100018815&fenleiid=4&sort=%b6%af%bb%ad&from=5&style=fullscreen&uid= 1&pid=h_home_inst&m=9e236bafe13c8348247781c2d0fab7a7&modulever=1.0.2.1025&appver=1.4.0.1040 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1; SV1; 4399) 9″Box.

58.50.201.130--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝103352&fenleiid＝10001&from＝6&leixing＝1&style＝iphone&uid＝1&pid＝h_home&m＝3d3e77348ff2fbfa6af7c3751a00edae&modulever =1.4.0.1040&appver=1.4.0.1040HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1)″

110.178.40.7--[20/Aug/2011:10:11:47+0800]″GET/stat.html？type＝open&action＝yingyongdianji&fangshi＝2&Appid＝100000525&fenleiid＝4&from＝0&leixing＝2&style＝fullscreen&uid＝1&pid＝softmgr&m＝f16b5a2c01d64fcfa3ff5f035ce74677&modulever =1.4.0.1041&appver=1.4.0.1041 HTTP/1.1″2000″-″″Mozilla/4.0(compatible; MSIE 7.0; Windows NT 5.1; Trident/4.0)″

58.50.201.130--[20/Aug/2011:10:11:47+0800] "GET/stat.html?type=open&action=zuixiaohuazhuomian&count=90468&uid=1&pid=h_home&m=3d3e77348ff2fbfa6af7c3751a00edae&modulever1=1.4.HTTP00. /1.1″200 0″-″″Mozilla/4.0(compatible; MSIE 6.0; Windows NT 5.1)″

The output data Data1 is:

00062ee80feec92758b8be8d3e4b9c16 100113997 2

0006d880a38f687c3fca10e8c4efa227 102019670 2

0007400ba8300fc80b7210a0e66de257 102022805 2

0007bda041324ee53b83f0343daf84d2 102005801 2

00080118032d30ad28a9bce8232d17c8 100013133 3

0008020587ea4fecf08ad58b09ef5904 110195914 2

00082bf0489199360bce8a06693ef3f5 100115004 3

 …

Map with the Appid_name comparison table to find out the application category corresponding to each Appid. The results are as follows:

00062ee80feec92758b8be8d3e4b9c16 1 100113997 2

0006d880a38f687c3fca10e8c4efa227 1 102019670 2

0007400ba8300fc80b7210a0e66de257 2 102022805 2

0007bda041324ee53b83f0343daf84d2 9 102005801 2

00080118032d30ad28a9bce8232d17c8 6 100013133 3

0008020587ea4fecf08ad58b09ef5904 6 110195914 2

00082bf0489199360bce8a06693ef3f5 4 100115004 3

 …

2). In the same Mid and feileid, sort the clicked or added apps according to the number of clicks or additions, take the top 20 Appids, and put them into one row.

Input data: Data1 (the output data of the previous step)

The data structure of the output data Data2 is: Mid fenleiid Appid1 weight1 Appid2weight2 Appid3 weight3......

For example: input data:

00082bf0489199360bce8a06693ef3f5 1 100115004 3

00083ebafe4eb71596f45dfa821f73d5 1 102028904 5

000887c0d3498c7c43cecb566a6333e4 8 102020030 3

0008ce3ae13aa6332794861b275861ad 10 102005157 2

 …

The output data Data2 is:

007a2663bac10378bcaa874be36a2d97 1 102006053 2 102028976 1100000913 1

007abe31b117a554df0fefc2a91200c2 2 120042762 11 102023358 2100000568 2 102010364 1 100115004 1

007b3dc8a31a6627a6e2468f789aa078  8  102007826  19 102020628  6102007664  6  102028968  4  100012183  3  100012315  3102043563  2  102022076  2  102000032  2  102031006  2110004672  2  102007377  2  101000009  2  110091072  2100030320  2  102044509  2  102043791  2  102044243  2102044665  2  100040423  2

007bb8043a487ed3a690aa6d461a3c10 10 102019572 3

007d072a6bfe28591f0eb4c5d533784c 18 100000525 31 10000028919 101000053 16 102005903  6  100000625  4  102020030  4102020628  3  120055003  3  100000913  3  102001686  2100034506  2  102005985  2  100000801  2  102044292  2110153628  2  100115575  2  100045261  2  100115650  2100103773  2  100102029  2

3). In the same pair of Mid and fenleiid, the Appid is paired in pairs, and the two Appids appear together once, and then the two Appids are used as a category to calculate the total number of times that the two Appids appear at the same time to generate a frequent 2-itemset.

Input data: Data2 (the output data of the previous step)

The data structure of the output data Data3 is: Appid1 Appid2 weight (number of occurrences)

Example: input data

Xx1 1 001 002 003 004...

Xx2 2 001 002 003...

Xx3 15 002 004...

 …

Intermediate result file:

001 002 1

001 003 1

001 004 1

002 003 1

002 004 1

003 004 1

001 002 1

002 005 1

002 004 1

The output data Data3 is:

001 002 2

001 003 1

001 004 1

002 003 2

002 004 2

003 004 1

4). Calculate the frequent 1-itemset, remember to calculate the number of occurrences of each Appid

Input data: Data1 (output data of step 1))

The data structure of the output data Data4 is: Appid weight (number of occurrences)

For example:

Input data:

Xx1 001 10

Xx1 002 8

Xx1 003 5

Xx2 002 8

Xx2 003 9

Xx3 003 7

 …

The output data Data4 is:

001 1

002 2

003 3

 …

5). Calculate confidence and sort by confidence

For example: the number of occurrences of application A and application B in the frequent 2-item set is N, and the number of occurrences of application A in the frequent 1-item set is M, then relative to A, the confidence of B is N/M, and the order is made according to the confidence , take the top 50 applications and combine them into one line with the first application as the class. The structure is:

A B C D  …

Input data: Data3 and Data4

The data structure of the output data Data5 is: Appid Appid1 weight1 Appid2weight2......Appid50 weight50

For example:

The input data Data3 is:

001 100 20

001 101 18

001 102 16

001 103 14

001 104 12

001 106 10

002 201 50

002 202 30

002 101 10

002 102 5

The input data Data4 is:

001 100

002 200

The output data Data5 is:

001 100 101 102 103 104 106...

002 201 201 101 102  …

6). According to Data1 and Data5, match according to user Appid to generate recommendation results.

For example:

Data1:

Xx1 1 001 10

Xx1 2 002 5

 …

Data5:

001 100 101 102 103 104 105 106 107...

002 200 201 202 203 204 205 206 207...

Match according to the underlined Appid, and generate the intermediate result as follows:

Xx1 1 100 101 102 103 104 105 106 107...10

Xx1 2 200 201 202 203 204 205 206 207...5

In the same pair of Mid and fenleiid, the top 50 applications are sorted by weight as follows:

Xx1 1 100 101 102 103 104 105 106 107...200 201 202 203 204 205206 207...

In a specific implementation, the results of Step 1 and Step 2 in the above example can be further merged according to Mid and fenleiid, filtered by the application category determined in step 103, and only the application data of the determined application category are taken, and then in each application category Randomly select 50 applications as the recommended results. If there is no data in a certain application category, you can use any mechanism to complete it, for example, put the most popular application, the latest application, etc.

Certainly, the above-mentioned method of finding an application that matches user behavior information is only used as an example, and it is also feasible for those skilled in the art to use other calculation methods, for example, by calculating the classification label of user behavior information and the classification label applied in the corresponding category application data set Matching degree, etc., the present application does not need to limit this.

Step 105: Generate a corresponding application folder according to the application category, and put the matched application into the corresponding application folder for recommendation.

Applying the embodiment of the present application, application folders will be generated by category, and applications under the corresponding category that match the user behavior information will be recommended to the user in the application folder of the corresponding category, thereby saving resources of the user device.

In a specific implementation, the application folders recommended to the user can be displayed in different split screens of the desktop. Preferably, the recommended application files in each split screen can also be determined according to the height and width of the user's split screen The number of clips. Applying the embodiment of this application, the display order of the application folders is set according to the operation frequency of each application category corresponding to the main classification label from high to low, so the application folders are displayed to the user according to the matching degree of user interests from high to low. In addition, the applications in the application folder are also sorted by weight, that is, they are displayed to the user according to the matching degree of the user's interest from high to low, so that the operation of the user is more convenient and the user can obtain a better experience.

In a specific implementation, icons corresponding to multiple application folders may be uniformly displayed on the user interface of the terminal desktop, each icon represents an application folder, and the icon is used as an application entry method. This graphical display is very intuitive for users, and it is easy to use and manage. For example, the icons displaying application folders in the user interface include "video", "novel", "education" and "game". Multiple application icons such as TV dramas, movies, animations, and variety shows are displayed in the sub-window. Using the icon as the application entry can prompt the user to use the application, but before the user actually chooses to use it, the configuration file corresponding to the application is not actually installed. Occupy more client resources.

The icons in the user interface can be centrally deployed or pushed by the central server on the network side, which prevents malicious programs from randomly adding malicious icons in the interface, further improving security. The configuration file managed centrally by the central server may include the access address of the corresponding application, the presentation specification, and the opening method of the application, or any combination thereof.

For example, for a web application, the web access address is sent by the central server to the terminal side through a configuration file, which prevents malicious programs on the terminal side from tampering with the access address.

Moreover, the central server on the network side can obtain updated configuration file information through interaction with a third-party content server. For example, if the access address of an application changes, the server will obtain updated address information through interaction with the content server, and Sending it through the configuration file eliminates the opportunity left by malicious programs due to the change of the access address.

In addition, after obtaining the configuration file of the application corresponding to the icon, the user equipment may also update the display state of the icon to further prompt the user. For example, before the profile is obtained, the icon can be black and white, or dark, and after obtaining it, it can be colored or bright.

It should also be noted that there may be one or more application folder icons displayed on the terminal-side user interface, which may be determined according to different display rules. For example, when one icon is used, the icon can be used as a unified entry for multiple lower-level applications or lower-level icons, and when any one of the applications obtains update information, a prompt can be obtained at the entry icon.

In a preferred embodiment of the present application, the following steps may also be included:

Collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID.

By establishing a user feature database, user behavior information can be processed uniformly on the server side or in the cloud. The application folder and the corresponding application that should be recommended to the user are determined based on the operation behavior information of the second time.

It should be noted that, for the method embodiment, for the sake of simple description, it is expressed as a series of action combinations, but those skilled in the art should know that the application is not limited by the described action sequence, because according to this application, certain steps may be performed in another order or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification belong to preferred embodiments, and the actions and modules involved are not necessarily required by this application.

Referring to FIG. 2 , it shows a structural block diagram of an embodiment of an application automatic recommendation device, which may specifically include the following modules:

A request receiving module 201, configured to receive an application acquisition request submitted by a user from a client, where the application acquisition request includes a user identifier;

The previous behavior information extraction module 202 is used to extract the existing user behavior information of the corresponding user from the user feature database according to the user identifier, and the user behavior information includes the user's operation information for the previous recommended application;

An application category determining module 203, configured to determine the application category recommended to the user according to the user behavior information;

A matching application acquisition module 204, configured to extract matching applications according to the user's operation information on previously recommended applications in the application data set of the application category;

The application recommendation module 205 is configured to generate a corresponding application folder according to the application category, and put the matching application into the corresponding application folder for recommendation.

In a specific implementation, the embodiment of the present application may also include the following modules:

The behavior statistics module is used to collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID.

As an example of a specific application of the embodiment of the present application, the user behavior information also includes the user's local operation behavior information, and/or, the user's online operation behavior information; in this case, the application category determination module 203 Can include the following submodules:

The first feature extraction submodule is used to extract classification labels and corresponding first operation frequency from the user's local operation behavior information and/or online operation behavior information;

A conversion submodule, configured to convert the classification label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of a classification label and an application category;

The second feature extraction sub-module is used to extract the application information and the corresponding second operation frequency operated by the user within the preset time period from the operation information of the user on the previously recommended application, the application information includes application category;

A sorting submodule, configured to calculate the weight of each application category according to the first operation frequency and the second operation frequency, and sort according to the weight of the application category from high to low;

The category selection sub-module is used to extract a preset number of top n application categories as application categories recommended to users; wherein, n is a positive integer greater than 1.

In a preferred embodiment of the present application, the device may also include the following modules:

The application data set generation module is used to generate application data sets of various application categories:

The application data set generation module is connected with the matching application acquisition module 204, and may specifically include the following submodules:

The similar application acquisition sub-module is used to acquire applications of the same application category, and the applications have classification tags;

The similarity calculation sub-module is used to determine the main application and the application to be recommended in the application, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application;

A quality scoring parameter acquisition submodule, configured to acquire the quality scoring parameters of the application to be recommended;

The application to be recommended extraction sub-module is used to extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the top m to be recommended Recommended applications; wherein, the m is a positive integer greater than 1;

The application data set forming sub-module is used to form the main application and the corresponding extracted application to be recommended into an application data set of the current application category.

In a preferred embodiment of the present application, the matching application acquisition module 204 may include the following submodules:

The main application statistics sub-module is used to count the main application and the corresponding third operation frequency according to the user's operation information on the previously recommended application, and the main application is an application operated by the user;

The application-to-be-recommended determination submodule is configured to extract a matching application to be recommended according to the main application in the application data set corresponding to the application category, and use the third operation frequency as the application in the matched application to be recommended A certain number of applications to be recommended are respectively extracted by the extracted weights, and applications to be recommended that satisfy the first preset number are extracted in total.

More preferably, the matching application acquisition module 204 may also include the following submodules:

The main application selection sub-module is used to obtain the application category corresponding to the main application, sort the main applications according to the third operation frequency within the same application category, and extract the first k main applications of the preset number; wherein, The k is a positive integer greater than 1;

The frequent 2-itemset calculation submodule is used to pair the extracted main applications in pairs, calculate the total number of simultaneous occurrences of the paired main applications, and generate frequent 2-itemsets;

The frequent 1-itemset calculation sub-module is used to calculate the number of occurrences of each main application and generate frequent 1-itemsets;

Confidence degree calculation sub-module, used to calculate the confidence degree of each main application according to the frequent 2-itemset and frequent 1-itemset, and sort the main applications according to the confidence degree;

The matching application determining submodule is configured to match the extracted applications to be recommended that meet the first preset number and the main applications sorted by confidence to generate a final recommended matching application.

The embodiments of the present application can be applied not only to an application environment of a single device, but also to a server-client application environment, or further to an application environment based on cloud technology.

Since the device embodiment basically corresponds to the foregoing method embodiment, for details not detailed in the description of this embodiment, reference may be made to the relevant description in the foregoing embodiment, and details are not repeated here. The modules, submodules and units involved in the device embodiments and system embodiments of the present application may be software, hardware, or a combination of software and hardware.

The application can be used in numerous general purpose or special purpose computing system environments or configurations. Examples: personal computers, server computers, handheld or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, including A distributed computing environment for any of the above systems or devices, etc.

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

Finally, it should also be noted that in this text, relational terms such as first and second etc. are only used to distinguish one entity or operation from another, and do not necessarily require or imply that these entities or operations, any such actual relationship or order exists. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

A method for automatic application recommendation and a device for automatic application recommendation provided by this application have been introduced in detail above. In this paper, specific examples have been used to illustrate the principle and implementation of this application. The description of the above embodiments is only It is used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation and application scope. In summary, this The content of the description should not be understood as limiting the application.

Claims (12)

1. A method for applying automatic recommendation, comprising: receiving an application acquisition request submitted by a user from a client, where the application acquisition request includes a user identifier; Extracting existing user behavior information of the corresponding user from the user feature database according to the user identifier, the user behavior information includes the user's operation information for the previously recommended application; Determine the application category recommended to the user according to the user behavior information; In the application data set of the application category, extract matching applications according to the user's operation information on previously recommended applications; A corresponding application folder is generated according to the application category, and the matching application is put into the corresponding application folder for recommendation. 2. The method of claim 1, further comprising: Collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID. 3. The method according to claim 1, wherein the user behavior information further includes the user's local operation behavior information, and/or, the user's online operation behavior information; The step of determining the application category recommended to the user according to the user behavior information includes: Extracting classification labels and corresponding first operation frequencies from the user's local operation behavior information and/or online operation behavior information; Converting the classification label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of the classification label and the application category; Extracting application information and a corresponding second operation frequency operated by the user within a preset time period from the user's operation information on previously recommended applications, the application information including application categories; Calculate the weight of each application category according to the first operation frequency and the second operation frequency, and sort according to the weight of the application category from high to low; A preset number of first n application categories are extracted as application categories recommended to the user; wherein, n is a positive integer greater than 1. 4. The method according to claim 3, wherein the application data set of a certain application category is generated through the following steps: Obtain applications of the same application category, where the applications have classification labels; Determine the main application and the application to be recommended in the application, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application; Obtaining quality scoring parameters of the application to be recommended; Extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the first m applications to be recommended by a preset number; wherein, m is A positive integer greater than 1; The main application and the extracted corresponding applications to be recommended are combined into an application data set of the current application category. 5. The method according to claim 4, wherein in the application data set of the application category, the step of extracting the matching application according to the operation information of the user for the previously recommended application comprises: According to the user's operation information on the previously recommended application, count the main application and the corresponding third operation frequency, the main application is the application operated by the user; In the application data set corresponding to the application category, the matching applications to be recommended are extracted according to the main application, and among the matched applications to be recommended, a certain number of applications to be recommended are respectively extracted using the third operation frequency as the weight of application extraction. For recommending applications, a total of applications to be recommended that satisfy the first preset number are extracted. 6. The method according to claim 5, wherein in the application data set of the application category, the step of extracting the matching application according to the operation information of the user for the previously recommended application further comprises: Obtain the application category corresponding to the main application, sort the main application according to the third operation frequency in the same application category, and extract the first k main applications of the preset number; wherein, the k is a positive number greater than 1 integer; Pairing the extracted main applications in pairs, calculating the total number of simultaneous occurrences of the paired main applications, and generating frequent 2-itemsets; Calculate the number of occurrences of each main application alone to generate frequent 1-itemsets; Calculate the confidence of each main application according to the frequent 2-itemset and frequent 1-itemset, and sort the main applications according to the confidence; Matching the extracted applications to be recommended satisfying the first preset number and the main applications sorted by confidence level to generate a final recommended matching application. 7. A device for automatically recommending applications, comprising: A request receiving module, configured to receive an application acquisition request submitted by a user from a client, where the application acquisition request includes a user identifier; The previous behavior information extraction module is used to extract the existing user behavior information of the corresponding user from the user feature database according to the user identification, and the user behavior information includes the user's operation information for the previous recommended application; An application category determination module, configured to determine the application category recommended to the user according to the user behavior information; The matching application acquisition module is used to extract matching applications according to the user's operation information on previously recommended applications in the application data set of the application category; The application recommendation module is configured to generate a corresponding application folder according to the application category, and put the matching application into the corresponding application folder for recommendation. 8. The device of claim 7, further comprising: The behavior statistics module is used to collect user behavior information after submitting the application acquisition request, and write it into the user feature database according to the user ID. 9. The device according to claim 7, wherein the user behavior information further includes the user's local operation behavior information, and/or, the user's online operation behavior information; The application category determination module includes: The first feature extraction submodule is used to extract classification labels and corresponding first operation frequency from the user's local operation behavior information and/or online operation behavior information; A conversion submodule, configured to convert the classification label into a corresponding application category according to a preset association rule; the preset association rule is a conversion rule of a classification label and an application category; The second feature extraction sub-module is used to extract the application information and the corresponding second operation frequency operated by the user within the preset time period from the operation information of the user on the previously recommended application, the application information includes application category; A sorting submodule, configured to calculate the weight of each application category according to the first operation frequency and the second operation frequency, and sort according to the weight of the application category from high to low; The category selection sub-module is used to extract a preset number of top n application categories as application categories recommended to users; wherein, n is a positive integer greater than 1. 10. The apparatus of claim 9, further comprising: The application data set generation module is used to generate application data sets of various application categories: specifically including: The similar application acquisition sub-module is used to acquire applications of the same application category, and the applications have classification tags; The similarity calculation sub-module is used to determine the main application and the application to be recommended in the application, and calculate the similarity between the application to be recommended and the main application according to the classification labels of each application; A quality scoring parameter acquisition submodule, configured to acquire the quality scoring parameters of the application to be recommended; The application to be recommended extraction sub-module is used to extract the applications to be recommended corresponding to the same main application, sort them according to the similarity and quality scoring parameters of each application to be recommended from high to low, and extract the top m to be recommended Recommended applications; wherein, the m is a positive integer greater than 1; The application data set forming sub-module is used to form the main application and the corresponding extracted application to be recommended into an application data set of the current application category. 11. The device according to claim 10, wherein the matching application acquiring module comprises: The main application statistics sub-module is used to count the main application and the corresponding third operation frequency according to the user's operation information on the previously recommended application, and the main application is an application operated by the user; The application-to-be-recommended determination submodule is configured to extract a matching application to be recommended according to the main application in the application data set corresponding to the application category, and use the third operation frequency as the application in the matched application to be recommended A certain number of applications to be recommended are respectively extracted by the extracted weights, and applications to be recommended that satisfy the first preset number are extracted in total. 12. The device according to claim 11, wherein the matching application acquisition module further comprises: The main application selection sub-module is used to obtain the application category corresponding to the main application, sort the main applications according to the third operation frequency within the same application category, and extract the first k main applications of the preset number; wherein, The k is a positive integer greater than 1; The frequent 2-itemset calculation submodule is used to pair the extracted main applications in pairs, calculate the total number of simultaneous occurrences of the paired main applications, and generate frequent 2-itemsets; The frequent 1-itemset calculation sub-module is used to calculate the number of occurrences of each main application and generate frequent 1-itemsets; Confidence degree calculation sub-module, used to calculate the confidence degree of each main application according to the frequent 2-itemset and frequent 1-itemset, and sort the main applications according to the confidence degree; The matching application determining submodule is configured to match the extracted applications to be recommended that meet the first preset number and the main applications sorted by confidence to generate a final recommended matching application.

Images