CN110569436A - network media news recommendation method based on high-dimensional auxiliary information - Google Patents
network media news recommendation method based on high-dimensional auxiliary information Download PDFInfo
- Publication number
- CN110569436A CN110569436A CN201910821595.XA CN201910821595A CN110569436A CN 110569436 A CN110569436 A CN 110569436A CN 201910821595 A CN201910821595 A CN 201910821595A CN 110569436 A CN110569436 A CN 110569436A
- Authority
- CN
- China
- Prior art keywords
- news
- user
- preference
- model
- vector
- 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
- 238000000034 method Methods 0.000 title claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 40
- 238000005457 optimization Methods 0.000 claims abstract description 25
- 238000004364 calculation method Methods 0.000 claims description 5
- 239000000126 substance Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3344—Query execution using natural language analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
the invention discloses a network media news recommendation method based on high-dimensional auxiliary information, which comprises the steps of generating main information characteristic vectors according to the titles and the abstracts of news; generating a high-dimensional auxiliary information feature vector according to the content of the news; establishing a preference prediction model of user news and an optimization model of prediction thereof; performing parameter optimization on the optimization model; bringing the optimized parameters into the preference prediction model; and recommending news to the user according to the scores calculated by the preference prediction model. The method fully utilizes the historical preference behavior of the specific user to provide information guidance for recommendation, the two feature vectors provide more comprehensive feature extraction for news, the minimum loss function is strived to achieve better fitting in a preference prediction model, and meanwhile sparse regularization punishment is introduced to prevent overfitting in the model optimization process, so that the method has better accuracy and quicker timeliness for the recommendation of the specific user.
Description
Technical Field
the invention belongs to the field of intelligent recommendation of text contents, and particularly relates to a network media news recommendation method based on high-dimensional auxiliary information.
background
in the age of rapid development of the internet, a great amount of novel information is gathered all the time. It is tedious and impractical for internet users to screen out the interesting, high-quality information by themselves. The search function is present to a certain extent to solve the information screening and recommendation problem, but is still far from sufficient. The search function requires that a user actively provide keywords to screen massive information, and when the user cannot accurately describe the own requirements, the screening effect of a search engine is greatly reduced. The appearance of personalized recommendation systems has greatly reduced the contradiction between the personalized needs of users and the huge data. The personalized recommendation technology recommends various resources required by the user for the user by researching the preference and interest of the user.
The network media news is an important means for enabling users to quickly know the world, and when the users perform automatic news recommendation, the users often have several problems, firstly, guidance on recommendation by using past behavior preference of specific users is utilized, secondly, recommendation of news is performed by only using a few keywords as characteristics and ignoring information brought by other texts, so that the recommendation effect is poor, and thirdly, whether news or articles just reported are recommended to the specific users is quickly judged. By properly solving the three problems, the personalized network media news can be more targeted and time-efficient.
Disclosure of Invention
the invention aims to provide a network media news recommending method based on high-dimensional auxiliary information, which is used for providing more targeted and time-efficient recommendation for a specific user for a network news media.
based on the above object, the invention provides a network media news recommendation method based on high-dimensional auxiliary information, which comprises the following steps:
step 1, acquiring a news set I and a preference vector R of a specific user to the news set;
step 2, generating a main information characteristic vector MF according to the headline and the abstract of the news; generating a high-dimensional auxiliary information feature vector FF according to the content of news;
Step 3, establishing a preference prediction model of user news and an optimization model of prediction thereof;
Step 4, optimizing parameters of the optimization model;
step 5, the optimized parameters are brought into the preference prediction model;
And 6, recommending news to the user according to the scores calculated by the preference prediction model.
the loss function in the optimization model for the user news prediction comprises two parts, namely loss of a preference prediction task and sparse regularization punishment of feature selection, wherein the loss of the preference prediction task is used for describing difference between a prediction result and an actual result, and the sparse regularization punishment of the feature selection is used for preventing overfitting in the model optimization process.
the optimization model of the user news prediction is as follows:
where n is the number of news, riIs the preference given by user u for news i;is the predicted preference of user u for news i, w1Weight vector, w, for user to main information feature vector of news2and alpha and beta are preset parameters for a weight vector of a user to a high-dimensional auxiliary information feature vector of news. Let x be a vector, xiAre elements of a vector, and their corresponding operators are respectively expressed as | | x1＝∑ixi|，||x2＝(∑ixi2)1/2。
The preference prediction model of the user news is as follows:
Wherein s isijIs the similarity between news i and j, R+a set of news representing the preferences of user u,j∈r + \ { i } represents that when the preference prediction of a user on a certain news is predicted, the existing preference information of the user on the news is not used;
s is as describedijThe calculation method comprises the following steps:Wherein, mfiIs the ith element of the FA and represents the ith characteristic attribute value, ffiIs the ith element of FF, representing the value of the ith characteristic attribute,. represents two vectors or the dot product of two vectors, and λ is the adjustable parameter.
In this way,
The parameter optimization comprises optimizing w by using random gradient descent1And w2Let us order
for calculationAnd
Iterative update w1：w2：
Stopping iteration if the maximum iteration times are reached or the error reaches a preset value;
Outputting the w after iterative optimization1And w2；
Wherein the content of the first and second substances,η is the step size.
And after the optimized parameters are obtained, calculating the preference score of the news to be recommended for the user according to the preference prediction model, and recommending and sequencing the news according to the score.
the invention relates to a network media news recommendation method based on high-dimensional auxiliary information, which fully utilizes the historical preference behavior of a specific user to provide information guidance for recommendation, and utilizes a main information characteristic vector and a high-dimensional auxiliary information characteristic vector to provide more comprehensive characteristic extraction for news so as to perform more accurate recommendation. The method has better accuracy and quicker timeliness for the recommendation of a specific user.
Drawings
fig. 1 is a schematic flowchart of a network media news recommendation method based on high-dimensional auxiliary information according to an embodiment of the present invention.
Detailed Description
the invention is further described with reference to the accompanying drawings, but the invention is not limited in any way, and any alterations or substitutions based on the teaching of the invention are within the scope of the invention.
Referring to fig. 1, an embodiment of the present invention is a method for recommending network media news based on high-dimensional auxiliary information, including the following steps:
step 1, acquiring a news set I and a preference vector R of a specific user to the news set;
Step 2, generating a main information characteristic vector MF according to the headline and the abstract of the news; generating a high-dimensional auxiliary information feature vector FF according to the content of news;
Step 3, establishing a preference prediction model of user news and an optimization model of prediction thereof;
step 4, optimizing parameters of the optimization model;
Step 5, the optimized parameters are brought into the preference prediction model;
And 6, recommending news to the user according to the scores calculated by the preference prediction model.
The loss function in the optimization model for the user news prediction comprises two parts, namely loss of a preference prediction task and sparse regularization punishment of feature selection, wherein the loss of the preference prediction task is used for describing difference between a prediction result and an actual result, and the sparse regularization punishment of the feature selection is used for preventing overfitting in the model optimization process.
The optimization model of the user news prediction is as follows:
where n is the number of news, riIs the preference given by user u for news i;is the predicted preference of user u for news i, w1Weight vector, w, for user to main information feature vector of news2Weighting vector of high-dimensional auxiliary information characteristic vector of news for user, alpha and beta are preset parameters, x is vector, x isiis the element of the vector, and the corresponding operators are respectively expressed as | | x | | luminance1＝∑i|xi|，||x||2＝(∑ixi 2)1/2。
The preference prediction model of the user news is as follows:
Wherein s isijIs the similarity between news i and j, R+A set of news representing the preferences of user u,j∈r + \ { i } represents that when the preference prediction of a user on a certain news is predicted, the existing preference information of the user on the news is not used;
S is as describedijthe calculation method comprises the following steps:wherein, mfiis the ith element of the FA and represents the ith characteristic attribute value, ffiIs the ith element of FF, representing the value of the ith characteristic attribute,. represents two vectors or the dot product of two vectors, and λ is the adjustable parameter.
In this way,
the parameter optimization comprises optimizing w by using random gradient descent1And w2Let us order
For calculationand
Iterative update w1：w2：
Stopping iteration if the maximum iteration times are reached or the error reaches a preset value;
Outputting the w after iterative optimization1And w2；
Wherein the content of the first and second substances,η is the step size.
And after the optimized parameters are obtained, calculating the preference score of the news to be recommended for the user according to the preference prediction model, and recommending and sequencing the news according to the score.
In conclusion, according to the network media news recommendation method based on the high-dimensional auxiliary information, the historical preference behavior of a specific user is fully utilized to provide information guidance for recommendation, meanwhile, the main information characteristic vector and the high-dimensional auxiliary information characteristic vector are utilized to provide more comprehensive characteristic extraction for news, so that more accurate recommendation is conducted, in a preference prediction model, the striving loss function is minimum to achieve better fitting, and meanwhile, sparse regularization punishment is introduced to prevent overfitting in the model optimization process. The method has better accuracy and quicker timeliness for the recommendation of a specific user.
Those of ordinary skill in the art will understand that: the discussion of any embodiment above is meant to be exemplary only, and is not intended to intimate that the scope of the disclosure, including the claims, is limited to these examples; within the idea of the invention, also features in the above embodiments or in different embodiments may be combined, steps may be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in detail for the sake of brevity. The embodiments of the invention are intended to embrace all such alternatives, modifications and variances that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, substitutions, improvements and the like that may be made without departing from the spirit and principles of the invention are intended to be included within the scope of the invention.
Claims (6)
1. A network media news recommendation method based on high-dimensional auxiliary information is characterized by comprising the following steps:
Step 1, acquiring a news set I and a preference vector R of a specific user to the news set;
step 2, generating a main information characteristic vector MF according to the headline and the abstract of the news; generating a high-dimensional auxiliary information feature vector FF according to the content of news;
Step 3, establishing a preference prediction model of user news and an optimization model of prediction thereof;
Step 4, optimizing parameters of the optimization model;
Step 5, the optimized parameters are brought into the preference prediction model;
And 6, recommending news to the user according to the scores calculated by the preference prediction model.
2. the method of claim 1, wherein the loss function in the optimization model for user news prediction comprises a loss of a preference prediction task and a sparse regularization penalty of feature selection, the loss of the preference prediction task is used for describing the difference between a prediction result and an actual result, and the sparse regularization penalty of feature selection is used for preventing overfitting in the model optimization process.
3. The method of claim 2, wherein the optimization model of the user news forecast is:
Where n is the number of news, riIs the preference given by user u for news i;is the predicted preference of user u for news i, w1weight vector, w, for user to main information feature vector of news2weighting vector of high-dimensional auxiliary information characteristic vector of news for user, alpha and beta are preset parameters, x is vector, x isiIs the element of the vector, and the corresponding operators are respectively expressed as | | x | | luminance1＝∑i|xi|，||x||2＝(∑ixi 2)1/2。
4. The method of claim 3, wherein the preference prediction model of the user news is:
Wherein s isijis the similarity between news i and j, R+A set of news representing the preferences of user u, j ∈ R+\ { i } represents that when predicting the preference of a user to a certain news, the existing preference information of the user to the news is not used;
S is as describedijThe calculation method comprises the following steps:Wherein, mfiIs the ith element of the FA and represents the ith characteristic attribute value, ffiIs the ith element of FF, representing the value of the ith characteristic attribute,. represents two vectors or the dot product of two vectors, and λ is the adjustable parameter.
5. The method of claim 4, wherein the optimizing of parameters in step 4 comprises optimizing w using a stochastic gradient descent1and w2Let us order
Step 401, calculateAnd
Step 402, iteratively updating w1：w2：
Step 403, stopping iteration if the maximum iteration times or the error reaches a preset value;
Step 404, outputting w after iterative optimization1and w2；
Wherein the content of the first and second substances,η is the step size.
6. the method for recommending network media news of claim 5, wherein after the optimized parameters are obtained, the preference score of the news to be recommended for the user is calculated according to the preference prediction model, and the news is recommended in a sorted manner according to the score.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910821595.XA CN110569436A (en) | 2019-09-02 | 2019-09-02 | network media news recommendation method based on high-dimensional auxiliary information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910821595.XA CN110569436A (en) | 2019-09-02 | 2019-09-02 | network media news recommendation method based on high-dimensional auxiliary information |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110569436A true CN110569436A (en) | 2019-12-13 |
Family
ID=68777267
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910821595.XA Pending CN110569436A (en) | 2019-09-02 | 2019-09-02 | network media news recommendation method based on high-dimensional auxiliary information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110569436A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111984867A (en) * | 2020-08-20 | 2020-11-24 | 北京奇艺世纪科技有限公司 | Network resource determination method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107103071A (en) * | 2017-04-21 | 2017-08-29 | 安徽大学 | News information classification method based on directly optimized PAUC algorithm |
| CN110033127A (en) * | 2019-03-14 | 2019-07-19 | 中国人民解放军国防科技大学 | Cold start project recommendation method based on embedded feature selection |
-
2019
- 2019-09-02 CN CN201910821595.XA patent/CN110569436A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107103071A (en) * | 2017-04-21 | 2017-08-29 | 安徽大学 | News information classification method based on directly optimized PAUC algorithm |
| CN110033127A (en) * | 2019-03-14 | 2019-07-19 | 中国人民解放军国防科技大学 | Cold start project recommendation method based on embedded feature selection |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111984867A (en) * | 2020-08-20 | 2020-11-24 | 北京奇艺世纪科技有限公司 | Network resource determination method and device |
| CN111984867B (en) * | 2020-08-20 | 2023-06-06 | 北京奇艺世纪科技有限公司 | Network resource determining method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110704739B (en) | Resource recommendation method and device and computer storage medium | |
| CN110457581B (en) | Information recommendation method and device, electronic equipment and storage medium | |
| CN110263244B (en) | Content recommendation method, device, storage medium and computer equipment | |
| CN110209922B (en) | Object recommendation method and device, storage medium and computer equipment | |
| CN108960719B (en) | Method and device for selecting products and computer readable storage medium | |
| CN110909182B (en) | Multimedia resource searching method, device, computer equipment and storage medium | |
| CN105335519A (en) | Model generation method and device as well as recommendation method and device | |
| CN112115377A (en) | Graph neural network link prediction recommendation method based on social relationship | |
| CN107577736B (en) | File recommendation method and system based on BP neural network | |
| CN107203558B (en) | Object recommendation method and device, and recommendation information processing method and device | |
| CN108804577B (en) | Method for estimating interest degree of information tag | |
| CN111639247A (en) | Method, apparatus, device and computer-readable storage medium for evaluating quality of review | |
| CN112836509A (en) | Expert system knowledge base construction method and system | |
| CN112632984A (en) | Graph model mobile application classification method based on description text word frequency | |
| CN111078858A (en) | Article searching method and device and electronic equipment | |
| CN111460819A (en) | Personalized comment text recommendation system and recommendation method based on fine-grained sentiment analysis | |
| CN112380451A (en) | Favorite content recommendation method based on big data | |
| CN112765484A (en) | Short video pushing method and device, electronic equipment and storage medium | |
| CN110110143B (en) | Video classification method and device | |
| Wei et al. | Online education recommendation model based on user behavior data analysis | |
| CN114840745A (en) | Personalized recommendation method and system based on graph feature learning and deep semantic matching model | |
| CN110019563B (en) | Portrait modeling method and device based on multi-dimensional data | |
| CN113449200B (en) | Article recommendation method and device and computer storage medium | |
| CN110569436A (en) | network media news recommendation method based on high-dimensional auxiliary information | |
| CN111931256B (en) | Color matching recommendation method, device, equipment and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191213 |
|
| RJ01 | Rejection of invention patent application after publication |