CN108647364A - A kind of prediction recommendation method based on mobile terminal application data - Google Patents
A kind of prediction recommendation method based on mobile terminal application data Download PDFInfo
- Publication number
- CN108647364A CN108647364A CN201810490042.6A CN201810490042A CN108647364A CN 108647364 A CN108647364 A CN 108647364A CN 201810490042 A CN201810490042 A CN 201810490042A CN 108647364 A CN108647364 A CN 108647364A
- Authority
- CN
- China
- Prior art keywords
- user
- mobile terminal
- matrix
- word
- prediction
- 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.)
- Granted
Links
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Information Transfer Between Computers (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of, and method is recommended in the prediction based on mobile terminal application data, the present invention uses mobile terminal application data update and supplement user property according to user, in conjunction with the advantages of topic model and mixing collaborative filtering prediction technique, the accuracy of recommendation is improved while in view of user individual, and the present invention can obtain the recessive preference of user according to user installation mobile terminal applicable cases, and the accuracy predicted using user's recessiveness preference is higher than the accuracy of the collaborative filtering based on user and the collaborative filtering based on article.
Description
Technical field
The present invention relates to mobile Internet fields, more particularly, to a kind of prediction based on mobile terminal application data
Recommendation method.
Background technology
Currently, mobile network user is creating and is sharing large scale network information, such as article of text, photo, video,
It is the important clue for excavating userspersonal information, individual subscriber essential information includes gender, the age, personal interest (for example,
Science and technology, amusement, sport), personal profession information (for example, researcher, student, software engineer, musician), personal orientation of emotion
(for example, optimistic, passive) etc..Our these personal information are referred to as user property.It is inferred to from users personal data
User property can play an important role in many aspects such as customer analysis, information retrieval, personalized recommendations.
User network behavioral data (webpage click amount, web page access duration etc.), service behavior data are currently mainly used
(webpage stay time, web page access depth etc.), user preference data (browsing/collection, comment content, life-form structure preference
Deng), user's buying behavior data (repurchase rate, the rates such as user's contribution) predict user property.But it is directed to network behavior
Classification prediction cannot all reach ideal effect, the accuracy rate of prediction remains unchanged not high.Recommend application only to user using store
Application more than recommended user's quantity, does not account for user property and users ' individualized requirement.
Invention content
The present invention is the defect overcome described in the above-mentioned prior art, further promotes the accurate of network behavior classification prediction
Rate provides a kind of prediction recommendation method based on mobile terminal application data.
In order to solve the above technical problems, technical scheme is as follows：
A kind of prediction recommendation method based on mobile terminal application data, includes the following steps：
S1. the initial data of acquisition customer mobile terminal application, cleans initial data, obtains mobile terminal application
Data；
S2. it crawls mobile terminal and applies store data, calculate the word each applied and be biased to；
S3. data are biased to according to the word that step S2 is obtained, calculate the recessive preference matrix of user；
S4. according to the recessive preference matrix of the obtained users of step S3, user's recessiveness matrix is calculated, builds user property meter
Calculate model；
S5. part is taken to predict user property with whole collaborative filtering；
S6. the prediction user property obtained according to step S5 is updated or supplements to original user information, using pushing away
It is that user recommends personalized application to recommend system.
Preferably, the step S1 is as follows：
S11. user's characteristic information is counted according to User ID, is distributed by each feature in the information of statistics unique
Digital ID；
S12. the exceptional value in user information is deleted；
S13. choose related coefficient often occurring, calculating feature and user property from treated data, extraction with
The big feature of user property related coefficient, removal and the small feature of user property related coefficient；
S14. counting user clicks the number of mobile terminal application, and user preference matrix is calculated by normalizing formula
Cn×m。
Preferably, the step S2 is as follows：
S21. it uses crawler capturing mobile phone to apply the application data in store, passes through natural language processing method, processing movement
Terminal applies word simultaneously distributes unique number ID；
S22. it converts using word vector w=(w the word of mobile terminal application description information to1,w2...wn),
In, wnIndicate n-th of the word occurred；
S23. all mobile terminal application vectors are combined the-word Matrix C that is appliedm×l, Matrix Cm×lIn include
The be described word of application, Matrix Cm×lIn value be 0 or 1；
S24. it uses TF-IDF algorithms to calculate the word each applied to be biased to, and TF-IDF values is replaced into application-word matrix
Value, obtain updated application-word Matrix Cm×l。
Preferably, the step S3 is as follows：
S31. according to Matrix Cn×mAnd Cm×lMultiplication obtains user-word Matrix Cn×l。
Preferably, the step S4 is as follows：
S41. user-word Matrix C is handled according to LDA topic modelsn×l, obtain user's latent subject Matrix Cn×tWith theme-
Word preference matrix Ct×l；
S42. according to Matrix Cn×tObtain user property CkWith the relationship between theme topic；
S43. known users attribute C is obtained according to NB Algorithmk, by formula P (topic | ck) obtain preference master
Inscribe topic；
S44. P (c are calculated according to NB Algorithmk| topic), the maximum P (c of select probabilityk| topic) value
As user property calculated value.
Preferably, the step S5 is as follows：
S51. improved Pearson came similarity ρ is utilizedu'vNeighbours' similarity calculating method, according to theme-word Matrix Ct×l
Calculate N number of neighbours and the user-theme Matrix C of themen×tCalculate the M neighbours of user；
S52. it calculates the collaborative filtering based on theme and obtains value uu,iWith the collaborative filtering value based on user
wu,i；
S53. it according to mixing collaborative filtering prediction algorithm, obtains following formula and calculates user property prediction result：
qu,i=α uu,i+(1-α)wu,i
Wherein, qu,iFor predicted value, α is coefficient of balance.
Preferably, the step S6 is as follows：
S61. user information is updated or supplemented according to the user property prediction result that step S53 is obtained；
S62. user property prediction result and user property are combined, commending system is that user recommends personalized mobile end
End application.
Compared with prior art, the advantageous effect of technical solution of the present invention is：
The present invention is by analyzing customer mobile terminal application data prediction user property, in conjunction with commending system, in view of
Improve the accuracy of recommendation while user individual, and the present invention can be with according to user installation mobile terminal applicable cases
The recessive preference of user is obtained, is increased significantly to prediction result accuracy using user's recessiveness preference prediction user property.
Description of the drawings
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technology description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
Obtain other attached drawings according to these attached drawings.
Fig. 1 is that method schematic diagram is recommended in the prediction based on mobile terminal application data.
Specific implementation mode
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent；
To those skilled in the art, it is to be appreciated that certain known features and its explanation, which may be omitted, in attached drawing
's.
The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
A kind of prediction recommendation method based on mobile terminal application data, includes the following steps：
S1. the initial data of acquisition customer mobile terminal application, cleans initial data, obtains mobile terminal application
Data；
S2. it crawls mobile terminal and applies store data, calculate the word each applied and be biased to；
S3. data are biased to according to the word that step S2 is obtained, calculate the recessive preference matrix of user；
S4. according to the recessive preference matrix of the obtained users of step S3, user's recessiveness matrix is calculated, builds user property meter
Calculate model；
S5. part is taken to predict user property with whole collaborative filtering；
S6. the prediction user property obtained according to step S5 is updated or supplements to original user information, using pushing away
It is that user recommends personalized application to recommend system.
In the present embodiment, step S1 is as follows：
S11. user's characteristic information is counted according to User ID, is distributed by each feature in the information of statistics unique
Digital ID；
S12. the exceptional value in user information is deleted；
S13. choose related coefficient often occurring, calculating feature and user property from treated data, extraction with
The big feature of user property related coefficient, removal and the small feature of user property related coefficient；
S14. counting user clicks the number of mobile terminal application, and user preference matrix is calculated by normalizing formula
Cn×m。
In the present embodiment, step S2 is as follows：
S21. it uses crawler capturing mobile phone to apply the application data in store, passes through natural language processing method, processing movement
Terminal applies word simultaneously distributes unique number ID；
S22. it converts using word vector w=(w the word of mobile terminal application description information to1,w2...wn),
In, wnIndicate n-th of the word occurred；
S23. all mobile terminal application vectors are combined the-word Matrix C that is appliedm×l, Matrix Cm×lIn include
The be described word of application, Matrix Cm×lIn value be 0 or 1；
S24. it uses TF-IDF algorithms to calculate the word each applied to be biased to, and TF-IDF values is replaced into application-word matrix
Value, obtain updated application-word Matrix Cm×l。
In the present embodiment, step S3 is as follows：
S31. according to Matrix Cn×mAnd Cm×lMultiplication obtains user-word Matrix Cn×l。
In the present embodiment, step S4 is as follows：
S41. user-word Matrix C is handled according to LDA topic modelsn×l, obtain user's latent subject Matrix Cn×tWith theme-
Word preference matrix Ct×l；
S42. according to Matrix Cn×tObtain user property CkWith the relationship between theme topic；
S43. known users attribute C is obtained according to NB Algorithmk, by formula P (topic | ck) obtain preference master
Inscribe topic；
S44. P (c are calculated according to NB Algorithmk| topic), the maximum P (c of select probabilityk| topic) value
As user property calculated value.
In the present embodiment, step S5 is as follows：
S51. improved Pearson came similarity ρ is utilizedu'vNeighbours' similarity calculating method, according to theme-word Matrix Ct×l
Calculate N number of neighbours and the user-theme Matrix C of themen×tCalculate the M neighbours of user；
S52. it calculates the collaborative filtering based on theme and obtains value uu,iWith the collaborative filtering value based on user
wu,i；
S53. it according to mixing collaborative filtering prediction algorithm, obtains following formula and calculates user property prediction result：
qu,i=α uu,i+(1-α)wu,i
Wherein, qu,iFor predicted value, α is coefficient of balance.
In the present embodiment, step S6 is as follows：
S61. user information is updated or supplemented according to the user property prediction result that step S53 is obtained；
S62. user property prediction result and user property are combined, commending system is that user recommends personalized mobile end
End application.
The concrete principle of the above method is as follows：
In the present invention program, as shown in Figure 1, obtaining user preference matrix first, acquisition customer mobile terminal is applied
Initial data cleans initial data, obtains the mobile terminal application data that can be used, according to User ID by user's point
The mobile phone hit is applied into statistics, and each unique number ID of mobile phone application distribution deletes the exceptional value in user property.
In specific implementation process, number normalization will click on.A bit using user's mobile phone of click as preference increase, frequent to click
It can be considered as and like the mobile phone application, user is built into key-value pair to the preference degree of each application, is finally converted into user's
Have a preference for matrix, the value of matrix corresponds to preference.
Mobile terminal application message is obtained using reptile, information is acquired using breadth-first strategy.It is answered collected
Natural language processing is carried out with information, extracts the word weight of each mobile terminal application.Mobile terminal is applied to each list
The unique number ID of word distribution, the word during mobile terminal is applied build key-value pair, and a mobile terminal application correspondence is multiple
Word, is converted into the vector of mobile terminal application, and vectorial value corresponds to respective weight.Mobile terminal application Vector Groups are closed
Come, forms the sparse application-word matrix of a higher-dimension.
In conjunction with user preference matrix and application-word matrix, user-word matrix is calculated.User preference matrix column number and
It is identical using-word matrix line number, each record of user preference matrix represents the preference that user applies mobile terminal
Distribution represents the tendency degree of word in mobile phone application using each record of-word matrix, two matrixes is combined,
Obtain tendency degree of the user to word.
User-word matrix is handled according to LDA topic models, obtains user-latent subject matrix and latent subject-word square
Battle array.User-theme matrix representative user is more deep close to user demand, theme-word square to the preference of certain class application theme
Battle array represents each theme and is made of which word.User property C is obtained according to user-theme matrixkWith the pass between theme topic
System.
NB Algorithm obtains known users attribute Ck, the calculation formula of preference theme topic be P (topic | ck)；
P (c are calculated according to NB Algorithmk| topic), the maximum P (c of select probabilityk| topic) value is as user property
Calculated value.
It takes part to predict user property with whole collaborative filtering, the N number of of theme is calculated according to theme-word matrix
Neighbours calculate the collaborative filtering based on theme and obtain value uu,i.The M neighbours of user are calculated according to user-theme matrix,
It calculates the collaborative filtering based on user and obtains value wu,i.Predicted value q is calculated using mixing collaborative filtering prediction algorithmu,i, add
Enter coefficient of balance α, obtains predictor formula qu,i=α uu,i+(1-α)wu,i。
Original user information is updated or supplemented according to the user property of prediction, the use of commending system is that user recommends individual character
Change application.
In conclusion the present invention usage behavior information applied according to customer mobile terminal predicts user property, it will be pre-
The user property of survey is updated and supplements to legacy data, while mixing collaborative filtering prediction algorithm proposed by the present invention,
The accuracy of prediction is improved while consideration user individual.
Wherein user property prediction technique extracts the recessive information of user using LDA topic models, recessive partially using user
Prediction user property increases significantly to prediction result accuracy well.
Wherein mixing collaborative filtering prediction algorithm has used the collaborative filtering based on user and based on theme, in conjunction with two
The advantages of kind algorithm, that is, the individual operation of user is considered, and improve the accuracy of prediction.
The shortcomings that being wherein directed to Pearson came calculating formula of similarity, proposes improved Pearson came calculating formula of similarity,
Formula can more calculate accurate result after improvement.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
All any modification, equivalent and improvement etc., should be included in the claims in the present invention made by within the spirit and principle of invention
Protection domain within.
Claims (7)
1. method is recommended in a kind of prediction based on mobile terminal application data, which is characterized in that include the following steps：
S1. the initial data of acquisition customer mobile terminal application, cleans initial data, obtains mobile terminal application number
According to；
S2. it crawls mobile terminal and applies store data, calculate the word each applied and be biased to；
S3. data are biased to according to the word that step S2 is obtained, calculate the recessive preference matrix of user；
S4. according to the recessive preference matrix of the obtained users of step S3, user's recessiveness matrix is calculated, structure user property calculates mould
Type；
S5. part is taken to predict user property with whole collaborative filtering；
S6. the prediction user property obtained according to step S5 is updated or supplements to original user information, is using recommendation
System is that user recommends personalized application.
2. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S1 is as follows：
S11. user's characteristic information is counted according to User ID, by each unique number of feature distribution in the information of statistics
Word ID；
S12. the exceptional value in user information is deleted；
S13. related coefficient often occurring, calculating feature and user property, extraction and user are chosen from treated data
The big feature of attribute related coefficient, removal and the small feature of user property related coefficient；
S14. counting user clicks the number of mobile terminal application, and user preference Matrix C is calculated by normalizing formulan×m。
3. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S2 is as follows：
S21. it uses crawler capturing mobile phone to apply the application data in store, by natural language processing method, handles mobile terminal
Using word and distribute unique number ID；
S22. it converts using word vector w=(w the word of mobile terminal application description information to1,w2...wn), wherein wnTable
N-th existing of word is shown；
S23. all mobile terminal application vectors are combined the-word Matrix C that is appliedm×l, Matrix Cm×lIn comprising application
Be described word, Matrix Cm×lIn value be 0 or 1；
S24. it uses TF-IDF algorithms to calculate the word each applied to be biased to, and TF-IDF values is replaced to the value of application-word matrix,
Obtain updated application-word Matrix Cm×l。
4. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S3 is as follows：
S31. according to Matrix Cn×mAnd Cm×lMultiplication obtains user-word Matrix Cn×l。
5. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S4 is as follows：
S41. user-word Matrix C is handled according to LDA topic modelsn×l, obtain user's latent subject Matrix Cn×tIt is inclined with theme-word
Good Matrix Ct×l；
S42. according to Matrix Cn×tObtain user property CkWith the relationship between theme topic；
S43. known users attribute C is obtained according to NB Algorithmk, by formula P (topic | ck) obtain preference theme
topic；
S44. P (c are calculated according to NB Algorithmk| topic), the maximum P (c of select probabilityk| topic) value conduct
User property calculated value.
6. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S5 is as follows：
S51. improved Pearson came similarity ρ is utilizedu'vNeighbours' similarity calculating method, according to theme-word Matrix Ct×lIt calculates
N number of neighbours of theme and user-theme Matrix Cn×tCalculate the M neighbours of user；
S52. it calculates the collaborative filtering based on theme and obtains value uu,iValue w is obtained with the collaborative filtering based on useru,i；
S53. it according to mixing collaborative filtering prediction algorithm, obtains following formula and calculates user property prediction result：
qu,i=α uu,i+(1-α)wu,i
Wherein, qu,iFor predicted value, α is coefficient of balance.
7. method is recommended in the prediction according to claim 1 based on mobile terminal application data, which is characterized in that the step
Rapid S6 is as follows：
S61. user information is updated or supplemented according to the user property prediction result that step S53 is obtained；
S62. user property prediction result and user property are combined, commending system is that user recommends personalized mobile terminal to answer
With.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810490042.6A CN108647364B (en) | 2018-05-21 | 2018-05-21 | Prediction recommendation method based on mobile terminal application data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810490042.6A CN108647364B (en) | 2018-05-21 | 2018-05-21 | Prediction recommendation method based on mobile terminal application data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108647364A true CN108647364A (en) | 2018-10-12 |
| CN108647364B CN108647364B (en) | 2021-10-29 |
Family
ID=63757364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810490042.6A Active CN108647364B (en) | 2018-05-21 | 2018-05-21 | Prediction recommendation method based on mobile terminal application data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108647364B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583949A (en) * | 2018-11-22 | 2019-04-05 | 中国联合网络通信集团有限公司 | A kind of user changes planes prediction technique and system |
| CN109726311A (en) * | 2018-12-21 | 2019-05-07 | 广州华多网络科技有限公司 | A kind of video recommendation method, device and equipment |
| CN109902229A (en) * | 2019-02-01 | 2019-06-18 | 中森云链(成都)科技有限责任公司 | A kind of interpretable recommended method based on comment |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831234A (en) * | 2012-08-31 | 2012-12-19 | 北京邮电大学 | Personalized news recommendation device and method based on news content and theme feature |
| CN103324690A (en) * | 2013-06-03 | 2013-09-25 | 焦点科技股份有限公司 | Mixed recommendation method based on factorization condition limitation Boltzmann machine |
| CN103886067A (en) * | 2014-03-20 | 2014-06-25 | 浙江大学 | Method for recommending books through label implied topic |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
| US9501762B2 (en) * | 2013-04-23 | 2016-11-22 | Dropbox, Inc. | Application recommendation using automatically synchronized shared folders |
| CN106897914A (en) * | 2017-01-25 | 2017-06-27 | 浙江大学 | A kind of Method of Commodity Recommendation and system based on topic model |
| CN107016058A (en) * | 2017-03-10 | 2017-08-04 | 浙江工业大学 | A kind of recommendation Forecasting Methodology based on attribute information preference self study |
-
2018
- 2018-05-21 CN CN201810490042.6A patent/CN108647364B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102831234A (en) * | 2012-08-31 | 2012-12-19 | 北京邮电大学 | Personalized news recommendation device and method based on news content and theme feature |
| US9501762B2 (en) * | 2013-04-23 | 2016-11-22 | Dropbox, Inc. | Application recommendation using automatically synchronized shared folders |
| CN103324690A (en) * | 2013-06-03 | 2013-09-25 | 焦点科技股份有限公司 | Mixed recommendation method based on factorization condition limitation Boltzmann machine |
| CN103886067A (en) * | 2014-03-20 | 2014-06-25 | 浙江大学 | Method for recommending books through label implied topic |
| CN104166668A (en) * | 2014-06-09 | 2014-11-26 | 南京邮电大学 | News recommendation system and method based on FOLFM model |
| CN106897914A (en) * | 2017-01-25 | 2017-06-27 | 浙江大学 | A kind of Method of Commodity Recommendation and system based on topic model |
| CN107016058A (en) * | 2017-03-10 | 2017-08-04 | 浙江工业大学 | A kind of recommendation Forecasting Methodology based on attribute information preference self study |
Non-Patent Citations (1)
| Title |
|---|
| 叶柏龙 等: "基于评分和项目特征的群组推荐方法", 《计算机应用研究》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109583949A (en) * | 2018-11-22 | 2019-04-05 | 中国联合网络通信集团有限公司 | A kind of user changes planes prediction technique and system |
| CN109726311A (en) * | 2018-12-21 | 2019-05-07 | 广州华多网络科技有限公司 | A kind of video recommendation method, device and equipment |
| CN109902229A (en) * | 2019-02-01 | 2019-06-18 | 中森云链(成都)科技有限责任公司 | A kind of interpretable recommended method based on comment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108647364B (en) | 2021-10-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110309427B (en) | Object recommendation method and device and storage medium | |
| CN111797321B (en) | Personalized knowledge recommendation method and system for different scenes | |
| Berjani et al. | A recommendation system for spots in location-based online social networks | |
| US9785888B2 (en) | Information processing apparatus, information processing method, and program for prediction model generated based on evaluation information | |
| CN109960761B (en) | Information recommendation method, device, equipment and computer readable storage medium | |
| CN107784066A (en) | Information recommendation method, device, server and storage medium | |
| CN111177559B (en) | Text travel service recommendation method and device, electronic equipment and storage medium | |
| CN110928993A (en) | User position prediction method and system based on deep cycle neural network | |
| US20150178265A1 (en) | Content Recommendation System using a Neural Network Language Model | |
| CN103793537A (en) | System for recommending individual music based on multi-dimensional time series analysis and achieving method of system | |
| CN105989074A (en) | Method and device for recommending cold start through mobile equipment information | |
| CN108629010B (en) | Web service recommendation method based on theme and service combination information | |
| CN105069172A (en) | Interest label generating method | |
| CN107222787A (en) | Video resource popularity prediction method | |
| CN108876058B (en) | News event influence prediction method based on microblog | |
| CN108647364A (en) | A kind of prediction recommendation method based on mobile terminal application data | |
| Wang et al. | Recurrent recommendation with local coherence | |
| CN113806630B (en) | Attention-based multi-view feature fusion cross-domain recommendation method and device | |
| CN108475256A (en) | Feature insertion is generated from homologous factors | |
| CN108133390A (en) | For predicting the method and apparatus of user behavior and computing device | |
| CN102957949A (en) | Device and method for recommending video to user | |
| CN114637911B (en) | Method for recommending next interest point of attention fusion perception network | |
| CN107885716A (en) | Text recognition method and device | |
| CN105678590A (en) | topN recommendation method for social network based on cloud model | |
| CN107368499B (en) | Client label modeling and recommending method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |