CN103578007A - Mixed recommendation system and method for intelligent device - Google Patents
Mixed recommendation system and method for intelligent device Download PDFInfo
- Publication number
- CN103578007A CN103578007A CN201210253651.2A CN201210253651A CN103578007A CN 103578007 A CN103578007 A CN 103578007A CN 201210253651 A CN201210253651 A CN 201210253651A CN 103578007 A CN103578007 A CN 103578007A
- Authority
- CN
- China
- Prior art keywords
- user
- application program
- candidate
- recommending data
- mixing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a mixed recommendation system and method for recommending application programs to an intelligent device. The mixed recommendation system based on a distributed architecture of a cloud platform comprises a log leading-in and analysis device, a user modeling device and a mixed recommending device, wherein the log leading-in and analysis device leads in an application program access log from a database, analyzes the leading-in application program access log and extracts information about the application program use situation, the user modeling device is used for creating an application program list comprising a plurality of application program items, using the information extracted by the log leading-in and analysis device, and conducting user modeling for users in a distributed mode based on the MapReduce programming model, and the mixed recommending device generates item-based first candidate recommendation data of the collaborative filtering algorithm, second candidate recommendation data recommended by Top-N, and content-based third candidate recommendation data of the collaborative filtering algorithm for each user, and then generates mixed recommendation data according to the preset sum of recommended items, the first proportion, the second proportion and the third proportion.
Description
Technical field
A kind of commending system and method that provides application program to recommend for smart machine is provided the application, relate in particular to a kind of in the distributed system architecture based on cloud platform, use the information of application program access log to generate respectively project-based collaborative filtering recommending data, content-based recommending data and the Top-N recommending data based on user's country origin, and according to predetermined ratio, the described recommending data generating is merged, thereby generate the technology of mixing recommending data.
Background technology
Along with scientific and technological development, having operating system, user, can freely to install and unload the smart machine of application program more and more, such as smart mobile phone, intelligent television, panel computer etc.Take TV as example, and present televisor has not only had the function of placing TV, and intelligent television can have open platform and carry operating system as computer.The program that user can surf the Net by intelligent television, plays games, free installation and the third party service provider such as uninstall or game provide.Because smart machine user's selection is more and more, user effectively determines that own interested application program is more and more difficult, requires a great deal of time and energy.Yet smart machine has recorded the nearly all usage log of user.We can be by finding the interested application program of each user to the analysis of daily record and mode excavation, then one group of interested application program of user's most probable is recommended to user, the problem of losing effectively to solve information overload and information, saves user's time.
The concept of recommending can be defined as: any personalized recommendation that can produce can effectively guide user to arrive the system of interested or useful object in personalized mode as output or in the very large space of possibility option.In ecommerce, commending system is exactly demand and the hobby of understanding and learn client, for user provides merchandise news and suggestion, recommends the interested commodity of its possibility, realizes the personalization of information service.
At present, conventional recommended technology has the recommendation of collaborative filtering, content-based recommendation, the recommendation based on effectiveness, the recommendation based on knowledge and rule-based recommendation etc.Above-mentioned every kind of recommended technology has its merits and demerits separately.For example, the recommendation of collaborative filtering can be processed complicated destructuring object, personalized recommendation data are provided, recommend quality can improve in time, help the advantages such as potential interest of finding user, but meanwhile, collaborative filtering is recommended in that to deal with Sparse Problem aspect, extendability aspect, new user's cold start-up problem aspect and new projects' cold start-up problem aspect etc. all undesirable.Again for example, the recommendation results that content-based recommended technology can provide intuitively, easily explain, there is not the cold start-up problem of new projects, also user's evaluating data etc. since not needing, but but can not effectively solve new user's cold start-up problem, the project of attribute complexity can not be solved well, and a large amount of historical data etc. need to be relied on.
When relating to the application program of recommending smart machine, also need to consider what time below:
1. smart machine is all new things to many users, uses the user of application program to occupy the minority in the user of current use intelligent television, and the item number that user evaluates only accounts for the part that project sum is very little, so need to consider Sparse Problem.
2. worldwide, all can increase a large amount of new users every day, although they do not have the historical record that uses, be necessary for them and produce recommendation, be i.e. new user's cold start-up problem.
3. there is now third party to participate in the exploitation of smart machine application program.After new application program is released, they must be recommended to the cold start-up problem of user ，Ji new projects.
4. the scale of the application program of smart machine is little a lot of with respect to the user of smart machine.For the efficiency of commending system, must consider data scale and extensibility.
5. the recommendation of personalization, to strengthen user's experience.
Therefore, need to a kind ofly for various users, provide commending system use, personalized recommendation.
Summary of the invention
The object of the present invention is to provide a kind of distributed earth to carry out the recommendation process of the application program of smart machine, and generate the mixing recommending data of the advantage that has merged multiple recommended technology, thereby can both generate the recommending data that correlativity is stronger for dissimilar user and new release/used application program.
According to an aspect of the present invention, provide a kind of for the mixing commending system of smart machine exemplary application program, the distributed structure/architecture of described mixing commending system based on cloud platform, comprise: daily record importing and resolver, from database, import application program access log, and based on MapReduce programming model, the application program access log importing is resolved, extract the information about application program behaviour in service, user modeling device, for creating the application program table that comprises a plurality of application program items, each application program item comprises application program identification and a plurality of attribute, described attribute comprises Application Type, payment type and at least one keyword, and for the information that usage log imports and resolver extracts, based on MapReduce programming model distributed earth, carry out user modeling for each user, for each user builds the user model of the preference data of at least one attribute that comprises user's application programs, mix recommended device, for information and the application program table of user modeling device establishment and each user's the user model that usage log imports and resolver extracts, be respectively each user and generate first candidate's recommending data of project-based collaborative filtering, be included in second candidate's recommending data of the application program that user belonging country frequency of utilization is high and the 3rd candidate's recommending data of content-based collaborative filtering, then according to predetermined recommendation entry sum and first, the second and the 3rd ratio, the first candidate's recommending data generating from each user respectively, second candidate's recommending data and the 3rd candidate's recommending data are chosen the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described user.
Mixing recommended device can comprise: the first recommended device, for the information of extracting from daily record importing and resolver based on MapReduce, obtain each user the frequency of utilization of each application program is inputted as project, according to predetermined project-based collaborative filtering, generate first candidate's recommending data for each user; The second recommended device, for the information that usage log imports and resolver extracts, generation is included in the countries use program listing of each application program access times of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user; The 3rd recommended device, for the application program table creating according to mixing recommended device, for original application program generating feature vector, the proper vector based on generating and each user's user model are calculated the similarity of each application program, and generate the 3rd candidate's recommending data for each user.
Described project-based collaborative filtering can be the project-based collaborative filtering of realizing in Apache Mahout machine learning storehouse.
Described first, second, and third ratio can be respectively 60%, 20% and 20%.
Described mixing commending system can also comprise: user api, for carrying out subscription authentication according to the authentication request of smart machine, the secured session of foundation and smart machine, the mixing recommending data during secured session, the user for smart machine being generated sends to described smart machine.
According to a further aspect in the invention, provide a kind of for the mixing recommend method of smart machine exemplary application program, described mixing recommend method comprises, in the distributed system architecture based on cloud platform: import application program access log from database, and the application program access log importing is resolved, extract the information about application program behaviour in service; Establishment comprises the application program table of a plurality of application program items, and each application program item comprises application program identification and a plurality of attribute, and described attribute comprises Application Type, payment type and at least one keyword; The information that use is extracted from application program access log, carries out user modeling based on MapReduce programming model distributed earth for each user, for each user's structure comprises the user model of preference data of at least one attribute of user's application programs; For using the information extracted from application program access log, the application program table of establishment and be the user model of each user's structure, be respectively each user generate project-based collaborative filtering first candidate's recommending data, be included in second candidate's recommending data of the application program that user belonging country frequency of utilization is high and the 3rd candidate's recommending data of content-based collaborative filtering; According to predetermined recommendation entry sum and first, second, and third ratio, from the first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data that generate for each user, choose respectively the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described each user.
Can be by carrying out concurrently following processing, for each user generates first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data: the information of extracting from application program access log based on MapReduce obtains each user the frequency of utilization of each application program is inputted as project, according to predetermined project-based collaborative filtering, generate first candidate's recommending data for each user; The information that use is extracted from application program access log, generation is included in the countries use program listing of each application program access times of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user; According to the application program table that mixes recommended device and create, be original application program generating feature vector, the proper vector based on generating and each user's user model are calculated the similarity of each application program, and generate the 3rd candidate's recommending data for each user.
Described project-based collaborative filtering can be the project-based collaborative filtering of realizing in Apache Mahout machine learning storehouse.
Described first, second, and third ratio can be respectively 60%, 20% and 20%.
Described mixing recommend method can also comprise: according to the authentication request of smart machine, carry out subscription authentication, and the secured session of foundation and smart machine, the mixing recommending data during secured session, the user for smart machine being generated sends to described smart machine.
Accompanying drawing explanation
By the description of carrying out below in conjunction with accompanying drawing, above and other object of the present invention and feature will become apparent, wherein:
Fig. 1 is the structured flowchart illustrating according to mixing commending system of the present invention;
Fig. 2 is the process flow diagram illustrating according to the overall operation of mixing recommend method of the present invention;
Fig. 3 illustrates the process flow diagram of carrying out user modeling according to mixing commending system of the present invention;
Fig. 4 illustrates mixing commending system according to the present invention according to project-based Collaborative Filtering Recommendation Algorithm, to generate the process flow diagram of first candidate's recommending data;
Fig. 5 illustrates mixing commending system according to the present invention according to Top-N algorithm, to generate the process flow diagram of second candidate's recommending data;
Fig. 6 illustrates mixing commending system according to the present invention to use content-based proposed algorithm to generate the process flow diagram of the 3rd candidate's recommending data;
Fig. 7 illustrates mixing commending system according to the present invention based on subscription authentication, to process the process flow diagram of user's request.
Embodiment
Below, with reference to accompanying drawing, describe embodiments of the invention in detail.
Of the present invention is the mixing commending system of the smart machine exemplary application program distributed structure/architecture based on cloud platform, and distributed earth is carried out the various recommendations that the present invention relates to and calculated and process, thereby improves processing speed and the efficiency of mass data.
Fig. 1 is the structured flowchart illustrating according to mixing commending system of the present invention.With reference to Fig. 1, according to of the present invention, comprise for the mixing commending system of smart machine exemplary application program: daily record importing and resolver 110, user modeling device 130, mix recommended device 140 and user API server 150.
Daily record importing and resolver 110 for example, import application program access log from (log server is stored) database, and based on MapReduce programming model, the application program access log importing is resolved, extract about application program behaviour in service information.
According to exemplary embodiment of the present invention, daily record importing and resolver 110 can use the information creating of extraction to comprise the Hive table cnt_log of the fields such as user ID (userId), application program identification (AppID), user belonging country (CountryID) and time (time).
According to exemplary embodiment of the present invention, daily record importing and resolver 110, when importing application program access log, copy to relevant journal file in temporary folder.Daily record importing and resolver 110, when resolving application program access log, deposit the information of described extraction or Hive table cnt_log under the file of appointment in.
The application program tables that the information that mixing recommended device 140 imports for usage log and resolver 110 extracts and user modeling device 130 create and the user model of structure generate respectively candidate's recommending data (first candidate's recommending data), Top-N candidate recommending data (second candidate's recommending data) and content-based candidate's recommending data (the 3rd candidate's recommending data) of project-based collaborative filtering.Afterwards, mix recommended device 140 according to predetermined recommendation entry sum and first, second, and third ratio, from the first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data that generate, choose respectively the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described user.According to a preferred embodiment of the invention, described first, second, and third ratio is respectively 60%, 20% and 20%, but described ratio can be adjusted as required.In addition can the part using described predetermined recommendation entry sum and first, second, and third ratio as system configuration information be stored in CONFIG.SYS.
According to exemplary embodiment of the present invention, mix recommended device 140 and comprise the first recommended device 141, the second recommended device 142 and the 3rd recommended device 143, it is respectively used to generate first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data.
The first recommended device 141 obtains each user for the information of extracting from daily record importing and resolver 110 based on MapReduce the frequency of utilization of each application program is inputted as project, according to predetermined project-based collaborative filtering, generates first candidate's recommending data for each user.
The information that the second recommended device 142 imports for usage log and resolver 110 extracts, generation is included in the countries use program listing of each application program access times of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user.
The 3rd recommended device 143 is for according to the application program table creating, for original application program generating feature vector, proper vector based on generating and each user's user model are calculated the similarity of each application program, and generate the 3rd candidate's recommending data for each user.
Fig. 2 is the process flow diagram illustrating according to the overall operation of mixing recommend method of the present invention.
With reference to Fig. 2, at step S210, the daily record importing and the resolver 110 that mix commending system import application program access log from database, and the application program access log importing is resolved, and extract the information about application program behaviour in service.At step S220, the user modeling device 130 that mixes commending system creates the application program table that comprises a plurality of application program items, each application program item comprises application program identification and a plurality of attribute, and described attribute comprises Application Type, payment type and at least one keyword.
According to exemplary embodiment of the present invention, described application program table is created as Hive Table A PP_INFO, this table can comprise sign, title, Application Type (Category), the payment type (Free/Paid) of application program, Keyword List etc.
Although according to exemplary embodiment of the present invention, create application program table by user modeling device 130, can, by other modules, carry out the operation of establishment application program table as mixed recommended device 140 or independent module.
At step S230, the user modeling device 130 that mixes commending system uses the information of extracting from application program access log, based on MapReduce programming model distributed earth, carry out user modeling for each user, for each user builds the user model of the preference data of at least one attribute that comprises user's application programs.
In step 240, the mixing recommended device 140 of mixing commending system is respectively each user by the first recommended device 141, the second recommended device 142 and the 3rd recommended device 143 and generates first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data.
At step S250, mix the mixing recommended device 140 of commending system according to predetermined recommendation entry sum and first, second, and third ratio, from the first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data that generate for each user, choose respectively the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described each user.Described first, second, and third ratio can be respectively 60%, 20% and 20%.
According to the embodiment that selects of the present invention, described project-based collaborative filtering adopts the project-based collaborative filtering of realizing in Apache Mahout machine learning storehouse.
Hereinafter with reference to Fig. 3, describe according to the modelling operability of the user modeling device 130 in mixing commending system of the present invention.Fig. 3 illustrates the process flow diagram of carrying out user modeling according to mixing commending system of the present invention.Mixing commending system of the present invention is usingd the information (" being designated hereinafter simply as log information ") extracted from application program access log as input, adopts the method for adding up, and based on MapReduce, realizes distributed user modeling.
With reference to Fig. 3, at step S242, each user's application program is used and added up.Be specially: the record that is application program from log information extraction event, the result of output " userID, appID, count " (being respectively user ID, application program identification, access times) form.For example: the key of Map function output is userID, and the value of output is appID; The output key of Reduce function is " userID, appID ", and output value is statistics number " count " (counting), and separator can be space or other specific characters.Then, based on above-mentioned Output rusults, create Hive table USER_APP_STA, this table can comprise three fields, is respectively userID, appID and count.
After this, at step S244, user modeling device is added up the preference of the payment type of each user's application programs.For example,, by the be applied statistics of program payment type of associated Hive Table A PP_INFO and USER_APP_STA, " userID, Free/Paid count " (user ID, the free/counting of paying); Then add up the preference of each user to the application program of Free (freely) and Paid (paying) type, for example, the key of Map function output is userID, the value of output is " Free count " or " Paid count ", and the output key of Reduce function is userID, output value is probability " Free prob " and " Paid prob ", and separator can be space.
At step S246, user modeling device is added up the preference of each user's application programs type (Category).Specifically, by the be applied statistics of Program Type of associated Hive Table A PP_INFO and USER_APP_STA, the field in result table is " userID, category, count " (user ID, Application Type, counting); Then, the preference of counting user to types of applications, for example, the output key of Map function is userID, output value is " category, count "; The output key of Reduce function is userID, and output value is " category, prob ", and separator can be space.
Then, at step S248, user modeling device is added up the preference of each user's application programs key word.Particularly, the be applied statistics of program key word of first associated Hive Table A PP_INFO and USER_APP_STA, in result table, field is " userID, keywords, count " (user ID, keyword, counting); Then the preference situation of counting user to each keyword (keyword).Such as: Map function output key is userID, and output value is " keyword, count "; The output key of Reduce function is userID, and output value is " keyword prob " (keyword, probability).
Finally, at step S249, user modeling device uses each user's who obtains from abovementioned steps information structuring user model, merges the preference of each user's application programs feature.Be specially, by Hive Table A PP_INFO statistics keyword list, by Hive Table A PP_INFO statistics category (Application Type) list, then for each user extracts the information that comprises Free/Paid, keyword, category, and according to log information structuring user's model.
Fig. 4 illustrates mixing commending system according to the present invention according to project-based Collaborative Filtering Recommendation Algorithm, to generate the process flow diagram of first candidate's recommending data.
The algorithm (Itembased Collaborative Filtering) that the first recommended device of exemplary embodiment of the present adopts Apache Mahout storehouse to realize carries out project-based collaborative filtering recommending, but design concept of the present invention is not limited to this, can adopt any other project-based collaborative filtering to generate first candidate's recommending data, at this, only describe the embodiment that uses Apache Mahout storehouse.
In the Collaborative Filtering Recommendation Algorithm in Apache Mahout storehouse, user (user) and project (item) are all expressed as long shaping, therefore when resolving application program access log and extracting user, project information, they need to be changed to growth shaping; After this algorithm generating recommendations, need again the data-switching that is represented long shaping to become common form (as character string).The concrete steps of following exemplary embodiment, the table name of wherein using, field name, field type, data mode etc. are all the forms adopting in the present embodiment, are not must be like this:
At step S2511, the first recommended device is carried out the input of the described proposed algorithm of operation and is prepared, namely from application program access log, extract each user to the frequency of utilization of each application program (service condition of user's application programs), and by user string conversion growth shaping wherein.Input shape is as " userIndex ItemID pref ", wherein, the user ID of the long of userIndex after for conversion, the sign that ItemID is application program, itself is exactly long.Pref is the preference value of this user to this project, is floating type data.
At step S2512, project-based Collaborative Filtering Recommendation Algorithm in the first recommended device operation Mahout storehouse, produces shape as the recommendation results of " userIndex, appList ", wherein in appList, has comprised the project of recommending and the preference value of prediction.
At step S2513, the user of application program is used in the first recommended device usage log Information Statistics, and generates shape as " userIndex, userID, null " data, the value that wherein userIndex is key, userID, the value that null is value.
At step S2515, the first recommended device converts described recommendation results to the form needing, and it can be converted in the present embodiment to the form of " userID, appList ".Particularly, first, recommendation results conversion is shaped as: the form of " userIndex, null, appList ", null wherein, the value that appList is value, the value that userIndex is key.Then, merge the data that above-mentioned steps and step S2513 generate, generate shape if " userID, appList " is as first candidate's recommending data, wherein, the value that userID is key, the value that appList is value.
Fig. 5 illustrates mixing commending system according to the present invention according to Top-N algorithm, to generate the process flow diagram of second candidate's recommending data.
The second recommended device of the present invention is used the situation of application program according to this state of national statistics user, generation is included in the countries use program listing of the access times of each application program of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user.Be below the concrete implementation step that generates according to an exemplary embodiment of the present invention second candidate's recommending data, table name wherein, field name, field type, data mode etc. are all the forms adopting in the present embodiment, are not must be like this:
With reference to Fig. 5, at step S2521, the access times of each application program of various countries in the second recommended device statistical log, the data that the form that generates every is " country, appID; count ", the data that generate are merged and sorted, obtain various countries and use the application list, and just sort by access times, create national the application list, as comprise the Hive table topn_country_apps of country STRING and apps STRING field.
At step S2523, usage log information, obtains the information of each user belonging country.Be specially, create the Hive table topn_up_log that comprises the fields such as userId STRING, country STRING, the national information based under topn_up_log extraction different user, and create Hive table topn_user_country.This table comprises two fields: user STRING, country STRING.
At step S2525, countries use program listing and the Information generation user candidate of user belonging country recommending data based on step S2521 and S2523 generation.Be specially, associated Hive table topn_country_apps and topn_user_country, to create new field user STRING, the Hive table topn_user_apps of apps STRING of comprising.
At step S2527, each user's of usage log Information Statistics application program is used historical data, obtains shape as the data of " userId, usedAppList ", wherein " usedAppList " can be " appID1^appID2^ ... " form.
After this, at step S2528, from the user candidate recommending data generating at step S2525, remove each user's who generates at step S2527 application program and use historical data, then from remaining the application list, choose several application programs above as second candidate's recommending data.
Fig. 6 illustrates mixing commending system according to the present invention to use content-based proposed algorithm to generate the process flow diagram of the 3rd candidate's recommending data.
The content description of content-based recommend method by computational item and the similarity between user model (profile) generate comprise predetermined quantity the most similar project as recommending data.Describe hereinbefore the establishment of user modeling device and comprised the application program table of a plurality of application program items, and be configured to the processing that each user builds user model.
Be below the concrete implementation step that exemplary embodiment according to the present invention generates second candidate's recommending data, use therein table name, field name, field type, data mode etc. are all the forms adopting in the present embodiment, rather than must be like this:
With reference to Fig. 6, at step S2531, the 3rd recommended device is obtained the information about the original application program of each user by associated Hive Table A pp_Info and cnt_log, for example, can build the Hive table cnt_unused_apps that comprises described information.
At step S2533, the 3rd recommended device is each attribute generating feature vector of the original application program of described each user.That is to say, generate and take the proper vector that for example attribute of Application Type, payment type, keyword is variable.
At step S2535, the 3rd recommended device is based on not using proper vector and the user model of application program to calculate similarity between the two.
For example, can use vectorial cosine method to calculate described similarity:
The angle that calculates gained is less, represents that described application program is more similar to user preference.Use the result of calculating to build and comprise that the HIVE of " userId, appID, similarity " (user ID, application program ID and similarity) shows app_sim_list.
Then, at step S2538, the 3rd recommended device merges the result obtained by step S2535, thereby generates the 3rd candidate's recommending data of the expression project the most similar to user preference for each user.
Referring to figs. 1 through Fig. 6, described according to mixing commending system of the present invention and mixed recommend method and use three kinds of recommend methods with different advantages generate candidate's recommending data and according to predetermined ratio, three kinds of candidate's recommending datas that generate merged above, thereby having generated mixing recommending data for each user.Described mixing commending system and the advantage of mixing the comprehensive three kinds of dissimilar proposed algorithms of recommend method, generation is for the recommending data of the different phase of different users, application program, thereby solve new user's cold start-up problem, the cold start-up problem of new projects, and the recommending data strong to End-user relevance is provided simultaneously.
Although represent with reference to preferred embodiment and described the present invention, it should be appreciated by those skilled in the art that in the situation that do not depart from the spirit and scope of the present invention that are defined by the claims, can carry out various modifications and conversion to these embodiment.
Claims (10)
1. be a mixing commending system for smart machine exemplary application program, the distributed structure/architecture of described mixing commending system based on cloud platform, comprising:
Daily record importing and resolver, import application program access log from database, and based on MapReduce programming model, the application program access log importing is resolved, and extracts the information about application program behaviour in service;
User modeling device, for creating the application program table that comprises a plurality of application program items, each application program item comprises application program identification and a plurality of attribute, described attribute comprises Application Type, payment type and at least one keyword, and for the information that usage log imports and resolver extracts, based on MapReduce programming model distributed earth, carry out user modeling for each user, for each user builds the user model of the preference data of at least one attribute that comprises user's application programs;
Mix recommended device, for information and the application program table of user modeling device establishment and each user's the user model that usage log imports and resolver extracts, be respectively each user and generate first candidate's recommending data of project-based collaborative filtering, be included in second candidate's recommending data of the application program that user belonging country frequency of utilization is high and the 3rd candidate's recommending data of content-based collaborative filtering, then according to predetermined recommendation entry sum and first, the second and the 3rd ratio, the first candidate's recommending data generating from each user respectively, second candidate's recommending data and the 3rd candidate's recommending data are chosen the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described user.
2. mixing commending system claimed in claim 1, wherein, mixes recommended device and comprises:
The first recommended device, for the information of extracting from daily record importing and resolver based on MapReduce, obtain each user the frequency of utilization of each application program is inputted as project, according to predetermined project-based collaborative filtering, generate first candidate's recommending data for each user;
The second recommended device, for the information that usage log imports and resolver extracts, generation is included in the countries use program listing of each application program access times of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user;
The 3rd recommended device, for the application program table creating according to mixing recommended device, for original application program generating feature vector, the proper vector based on generating and each user's user model are calculated the similarity of each application program, and generate the 3rd candidate's recommending data for each user.
3. mixing commending system as claimed in claim 2, wherein, described project-based collaborative filtering is the project-based collaborative filtering of realizing in Apache Mahout machine learning storehouse.
4. mixing commending system as claimed in claim 1, wherein, described first, second, and third ratio is respectively 60%, 20% and 20%.
5. mixing commending system as claimed in claim 1, also comprises:
User api, for carrying out subscription authentication according to the authentication request of smart machine, the secured session of foundation and smart machine, the mixing recommending data during secured session, the user for smart machine being generated sends to described smart machine.
6. be a mixing recommend method for smart machine exemplary application program, described mixing recommend method comprises, in the distributed system architecture based on cloud platform:
From database, import application program access log, and the application program access log importing is resolved, extract the information about application program behaviour in service;
Establishment comprises the application program table of a plurality of application program items, and each application program item comprises application program identification and a plurality of attribute, and described attribute comprises Application Type, payment type and at least one keyword;
The information that use is extracted from application program access log, carries out user modeling based on MapReduce programming model distributed earth for each user, for each user's structure comprises the user model of preference data of at least one attribute of user's application programs;
For using the information extracted from application program access log, the application program table of establishment and be the user model of each user's structure, be respectively each user generate project-based collaborative filtering first candidate's recommending data, be included in second candidate's recommending data of the application program that user belonging country frequency of utilization is high and the 3rd candidate's recommending data of content-based collaborative filtering;
According to predetermined recommendation entry sum and first, second, and third ratio, from the first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data that generate for each user, choose respectively the recommending data entry of the quantity that meets corresponding proportion, and the recommending data entry of choosing is respectively merged to the mixing recommending data as described each user.
7. mixing recommend method as claimed in claim 6 wherein, is processed below carrying out concurrently, for each user generates first candidate's recommending data, second candidate's recommending data and the 3rd candidate's recommending data:
The information of extracting from application program access log based on MapReduce obtains each user the frequency of utilization of each application program is inputted as project, according to predetermined project-based collaborative filtering, generates first candidate's recommending data for each user;
The information that use is extracted from application program access log, generation is included in the countries use program listing of each application program access times of every country, obtain the information of each user belonging country, add up each user and use the historical data of application program, and generate second candidate's recommending data based on above-mentioned information for each user;
According to the application program table that mixes recommended device and create, be original application program generating feature vector, the proper vector based on generating and each user's user model are calculated the similarity of each application program, and generate the 3rd candidate's recommending data for each user.
8. mixing recommend method as claimed in claim 7, described project-based collaborative filtering is the project-based collaborative filtering of realizing in Apache Mahout machine learning storehouse.
9. mixing recommend method as claimed in claim 6, wherein, described first, second, and third ratio is respectively 60%, 20% and 20%.
10. mixing recommend method as claimed in claim 6, also comprise: according to the authentication request of smart machine, carry out subscription authentication, the secured session of foundation and smart machine, the mixing recommending data during secured session, the user for smart machine being generated sends to described smart machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210253651.2A CN103578007A (en) | 2012-07-20 | 2012-07-20 | Mixed recommendation system and method for intelligent device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210253651.2A CN103578007A (en) | 2012-07-20 | 2012-07-20 | Mixed recommendation system and method for intelligent device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103578007A true CN103578007A (en) | 2014-02-12 |
Family
ID=50049742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210253651.2A Pending CN103578007A (en) | 2012-07-20 | 2012-07-20 | Mixed recommendation system and method for intelligent device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103578007A (en) |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104104729A (en) * | 2014-07-24 | 2014-10-15 | 福州瑞芯微电子有限公司 | General setting method and system for mobile equipment |
| CN104486648A (en) * | 2014-12-15 | 2015-04-01 | 四川长虹电器股份有限公司 | Recommended method and server |
| CN105204356A (en) * | 2015-09-17 | 2015-12-30 | 小米科技有限责任公司 | Application display method, device and terminal |
| CN105472460A (en) * | 2015-11-27 | 2016-04-06 | 广东长虹电子有限公司 | Intelligent sorting method and system of television applications |
| CN106503025A (en) * | 2015-09-08 | 2017-03-15 | 北京搜狗科技发展有限公司 | Method and system is recommended in a kind of application |
| CN106503224A (en) * | 2016-11-04 | 2017-03-15 | 维沃移动通信有限公司 | A kind of method and device for recommending application according to keyword |
| CN106610970A (en) * | 2015-10-21 | 2017-05-03 | 上海文广互动电视有限公司 | Collaborative filtering-based content recommendation system and method |
| CN107169572A (en) * | 2016-12-23 | 2017-09-15 | 福州大学 | A kind of machine learning Service Assembly method based on Mahout |
| CN107391687A (en) * | 2017-07-24 | 2017-11-24 | 华中师范大学 | A kind of mixing commending system towards local chronicle website |
| CN107506488A (en) * | 2017-09-21 | 2017-12-22 | 程丹秋 | A kind of intelligent music commending system |
| WO2018036307A1 (en) * | 2016-08-23 | 2018-03-01 | 腾讯科技（深圳）有限公司 | Information processing method utilized in pushing information order, allocation method, device, and data storage medium |
| CN108011941A (en) * | 2017-11-29 | 2018-05-08 | 广东欧珀移动通信有限公司 | Content delivery method, device, server and storage medium |
| CN108520017A (en) * | 2018-03-21 | 2018-09-11 | 广东欧珀移动通信有限公司 | Application program recommends method, apparatus, server and storage medium |
| CN108960986A (en) * | 2018-06-26 | 2018-12-07 | 西安交通大学 | A kind of supplier's recommended method based on web crawlers |
| CN109299349A (en) * | 2018-09-11 | 2019-02-01 | 广州视源电子科技股份有限公司 | Application recommendation method and device, equipment and computer-readable storage medium |
| CN109324955A (en) * | 2018-09-14 | 2019-02-12 | 科大国创软件股份有限公司 | A kind of IT operational system interface creating method with intelligent recommendation function |
| CN109688458A (en) * | 2019-01-14 | 2019-04-26 | 四川长虹电器股份有限公司 | The implementation method of smart television cloud desktop operation system based on big data algorithm |
| CN109903107A (en) * | 2017-12-08 | 2019-06-18 | 北京京东尚科信息技术有限公司 | Method and apparatus for generating information |
| CN109977316A (en) * | 2019-03-29 | 2019-07-05 | 厦门铠甲网络股份有限公司 | A kind of parallel type article recommended method, device, equipment and storage medium |
| CN110209931A (en) * | 2019-05-17 | 2019-09-06 | 腾讯科技（深圳）有限公司 | Method for pushing and device, storage medium, the electronic device of media content |
| CN111460271A (en) * | 2019-01-19 | 2020-07-28 | 上海易慕网络科技有限公司 | Intelligent personalized recommendation algorithm for network information |
| WO2021012790A1 (en) * | 2019-07-22 | 2021-01-28 | 平安科技（深圳）有限公司 | Page data generation method and apparatus, computer device, and storage medium |
| CN113744021A (en) * | 2021-02-08 | 2021-12-03 | 北京沃东天骏信息技术有限公司 | Recommendation method, recommendation device, computer storage medium and recommendation system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859425A (en) * | 2010-06-02 | 2010-10-13 | 中兴通讯股份有限公司 | Method and device for providing application list |
| US20110131077A1 (en) * | 2009-12-01 | 2011-06-02 | Microsoft Corporation | Context-Aware Recommendation Module Using Multiple Models |
| CN102236669A (en) * | 2010-04-30 | 2011-11-09 | 腾讯科技（深圳）有限公司 | Recommendation list generation method, device, media server, client and method |
| CN102360364A (en) * | 2011-09-30 | 2012-02-22 | 奇智软件（北京）有限公司 | Automatic application recommendation method and device |
-
2012
- 2012-07-20 CN CN201210253651.2A patent/CN103578007A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110131077A1 (en) * | 2009-12-01 | 2011-06-02 | Microsoft Corporation | Context-Aware Recommendation Module Using Multiple Models |
| CN102236669A (en) * | 2010-04-30 | 2011-11-09 | 腾讯科技（深圳）有限公司 | Recommendation list generation method, device, media server, client and method |
| CN101859425A (en) * | 2010-06-02 | 2010-10-13 | 中兴通讯股份有限公司 | Method and device for providing application list |
| CN102360364A (en) * | 2011-09-30 | 2012-02-22 | 奇智软件（北京）有限公司 | Automatic application recommendation method and device |
Non-Patent Citations (4)
| Title |
|---|
| 公丽萍: "个性化书籍推荐系统研究与设计", 《中国优秀硕士学位论文全文数据库信息科技辑》, vol. 2012, no. 05, 15 May 2012 (2012-05-15), pages 138 - 1637 * |
| 林立宇等: "基于云计算的电子商务推荐平台的构建分析", 《广东通信技术》, vol. 30, no. 11, 30 November 2010 (2010-11-30), pages 7 - 10 * |
| 马丽丽: "MapReduce编程模型在日志分析方面的应用", 《HTTP://WWW.IBM.COM/DEVELOPERWORKS/CN/JAVA/JAVA-LO-MAPREDUCE/》, 5 May 2011 (2011-05-05) * |
| 马春娥等: "基于Apache Mahout构建社会化推荐引擎", 《HTTP://WWW.IBM.COM/DEVELOPERWORKS/CN/JAVA/J-LO-MAHOUT/》, 21 January 2010 (2010-01-21) * |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104104729B (en) * | 2014-07-24 | 2017-10-27 | 福州瑞芯微电子股份有限公司 | The general method to set up of mobile device and system |
| CN104104729A (en) * | 2014-07-24 | 2014-10-15 | 福州瑞芯微电子有限公司 | General setting method and system for mobile equipment |
| CN104486648A (en) * | 2014-12-15 | 2015-04-01 | 四川长虹电器股份有限公司 | Recommended method and server |
| CN106503025A (en) * | 2015-09-08 | 2017-03-15 | 北京搜狗科技发展有限公司 | Method and system is recommended in a kind of application |
| CN105204356A (en) * | 2015-09-17 | 2015-12-30 | 小米科技有限责任公司 | Application display method, device and terminal |
| CN105204356B (en) * | 2015-09-17 | 2018-09-04 | 小米科技有限责任公司 | Display methods, device and the terminal of application |
| CN106610970A (en) * | 2015-10-21 | 2017-05-03 | 上海文广互动电视有限公司 | Collaborative filtering-based content recommendation system and method |
| CN105472460A (en) * | 2015-11-27 | 2016-04-06 | 广东长虹电子有限公司 | Intelligent sorting method and system of television applications |
| WO2018036307A1 (en) * | 2016-08-23 | 2018-03-01 | 腾讯科技（深圳）有限公司 | Information processing method utilized in pushing information order, allocation method, device, and data storage medium |
| CN106503224A (en) * | 2016-11-04 | 2017-03-15 | 维沃移动通信有限公司 | A kind of method and device for recommending application according to keyword |
| CN106503224B (en) * | 2016-11-04 | 2019-09-27 | 维沃移动通信有限公司 | A kind of method and device for recommending application according to keyword |
| CN107169572A (en) * | 2016-12-23 | 2017-09-15 | 福州大学 | A kind of machine learning Service Assembly method based on Mahout |
| CN107391687A (en) * | 2017-07-24 | 2017-11-24 | 华中师范大学 | A kind of mixing commending system towards local chronicle website |
| CN107391687B (en) * | 2017-07-24 | 2020-04-03 | 华中师范大学 | Local log website-oriented hybrid recommendation system |
| CN107506488A (en) * | 2017-09-21 | 2017-12-22 | 程丹秋 | A kind of intelligent music commending system |
| CN108011941A (en) * | 2017-11-29 | 2018-05-08 | 广东欧珀移动通信有限公司 | Content delivery method, device, server and storage medium |
| CN109903107A (en) * | 2017-12-08 | 2019-06-18 | 北京京东尚科信息技术有限公司 | Method and apparatus for generating information |
| CN108520017A (en) * | 2018-03-21 | 2018-09-11 | 广东欧珀移动通信有限公司 | Application program recommends method, apparatus, server and storage medium |
| CN108960986A (en) * | 2018-06-26 | 2018-12-07 | 西安交通大学 | A kind of supplier's recommended method based on web crawlers |
| CN109299349A (en) * | 2018-09-11 | 2019-02-01 | 广州视源电子科技股份有限公司 | Application recommendation method and device, equipment and computer-readable storage medium |
| CN109324955A (en) * | 2018-09-14 | 2019-02-12 | 科大国创软件股份有限公司 | A kind of IT operational system interface creating method with intelligent recommendation function |
| CN109688458A (en) * | 2019-01-14 | 2019-04-26 | 四川长虹电器股份有限公司 | The implementation method of smart television cloud desktop operation system based on big data algorithm |
| CN111460271A (en) * | 2019-01-19 | 2020-07-28 | 上海易慕网络科技有限公司 | Intelligent personalized recommendation algorithm for network information |
| CN109977316A (en) * | 2019-03-29 | 2019-07-05 | 厦门铠甲网络股份有限公司 | A kind of parallel type article recommended method, device, equipment and storage medium |
| CN110209931A (en) * | 2019-05-17 | 2019-09-06 | 腾讯科技（深圳）有限公司 | Method for pushing and device, storage medium, the electronic device of media content |
| CN110209931B (en) * | 2019-05-17 | 2023-08-25 | 腾讯科技（深圳）有限公司 | Media content pushing method and device, storage medium and electronic device |
| WO2021012790A1 (en) * | 2019-07-22 | 2021-01-28 | 平安科技（深圳）有限公司 | Page data generation method and apparatus, computer device, and storage medium |
| CN113744021A (en) * | 2021-02-08 | 2021-12-03 | 北京沃东天骏信息技术有限公司 | Recommendation method, recommendation device, computer storage medium and recommendation system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103578007A (en) | Mixed recommendation system and method for intelligent device | |
| CN106296059B (en) | Method and equipment for determining delivery network points | |
| CN110705683B (en) | Random forest model construction method and device, electronic equipment and storage medium | |
| CN110688476B (en) | Text recommendation method and device based on artificial intelligence | |
| CN103488678B (en) | Friend recommendation system based on user sign-in similarity | |
| CN102567511B (en) | Method and device for automatically recommending application | |
| Chen | Personalized recommendation system of e-commerce based on big data analysis | |
| JP2020509449A (en) | Method and device for warning | |
| CN102999586B (en) | A kind of method and apparatus of recommendation of websites | |
| CN104063801A (en) | Mobile advertisement recommendation method based on cluster | |
| US20140282493A1 (en) | System for replicating apps from an existing device to a new device | |
| CN106294787A (en) | Information pushing method and device and electronic equipment | |
| CN103106285A (en) | Recommendation algorithm based on information security professional social network platform | |
| CN111026858B (en) | Project information processing method and device based on project recommendation model | |
| CN105989074A (en) | Method and device for recommending cold start through mobile equipment information | |
| CN103473230A (en) | Service range determining method, logistics service provider recommending method and corresponding device | |
| CN103455559A (en) | Method and device for automatically recommending application | |
| CN103488788A (en) | Method and device for recommending applications automatically | |
| CN102541853A (en) | Method and device which are capable of obtaining application information by utilizing browser address bar | |
| CN102939774A (en) | Method and apparatus for context-indexed network resource sections | |
| WO2012102761A1 (en) | Method and system for providing detailed information in an interactive manner in a short message service (sms) environment | |
| CN104751354A (en) | Advertisement cluster screening method | |
| CN109074366A (en) | Gain adjustment component for computer network routed infrastructure | |
| Hasbolah et al. | A review on cyber entrepreneurship in Malaysia: past, present and future | |
| CN104484435A (en) | Method for cross-over analysis of user behavior |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20140212 |