CN114268836A - Cold start recommendation method and system for television applet - Google Patents
Cold start recommendation method and system for television applet Download PDFInfo
- Publication number
- CN114268836A CN114268836A CN202111356474.6A CN202111356474A CN114268836A CN 114268836 A CN114268836 A CN 114268836A CN 202111356474 A CN202111356474 A CN 202111356474A CN 114268836 A CN114268836 A CN 114268836A
- Authority
- CN
- China
- Prior art keywords
- user
- television
- data
- applet
- embedding
- 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)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
Abstract
The invention provides a cold start recommendation method and system for a television applet, which comprises the following steps: step 1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode; step 2: preprocessing the collected behavior data; and step 3: performing machine learning training by using a factor decomposition machine FM algorithm to obtain user data, and judging the similarity of users according to the distance; and 4, step 4: clustering user data by using a K mean value clustering algorithm; and 5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user. The invention solves the problems that the data volume of the user at the current stage of the television small program is small and the recommendation is difficult to be made.
Description
Technical Field
The invention relates to the technical field of data recommendation, in particular to a cold start recommendation method and system for a television applet.
Background
The television applet is a product in a development stage on a television, the user quantity is small, accurate recommendation is difficult to achieve for each user, and the relative watching behaviors of the user on the television are large.
Patent document CN112528164A (application number: CN202011465765.4) discloses a method and apparatus for user collaborative filtering recall, however, this patent scenario requires that users must act on products to recommend them, and for a new product, not every user acts on it, so that it is not suitable for the recommendation of such products
Patent document CN107590245A (application number: CN201710829260.3) discloses a light application recommendation method, device and electronic device, however, the reality required by the patent is unique to the mobile terminal, and is not applicable to a fixed product such as a television.
Disclosure of Invention
In view of the defects in the prior art, the invention aims to provide a cold start recommendation method and system for a television applet.
The cold start recommendation method for the television small program provided by the invention comprises the following steps:
step 1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode;
step 2: preprocessing collected behavior data, including data cleaning, marking, merging and weighting;
and step 3: performing machine learning training by using a factor decomposition machine FM algorithm according to the preprocessed data to obtain User data User Embedding, and judging the similarity of users according to the distance between the User Embedding;
and 4, step 4: based on the distance between the User embeddings, clustering the User embeddings by using a K-means clustering algorithm;
and 5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user.
Preferably, the behavior data using movie on demand includes: behaviors of watching, collecting and canceling video contents, and names, duration and types of videos;
the behavior data using the television applet includes: the behavior of using, collecting and canceling the small program, and the name, the using time length and the type of the small program;
and storing the behavior data using the film and television on demand and the behavior data using the television small program in a database.
Preferably, for behavior data requested by using movies and televisions, marking the data according to the operation behavior and watching duration indexes of a user to represent the interest of the movies and televisions, sampling and combining the data according to the interest degree of the user to the movies and televisions according to a preset proportion to form a training data set;
for the behavior data using the television applet, the behaviors of using, collecting and canceling the collection are respectively given weight, and the favorite applet of each user is calculated.
Preferably, the formula of the FM algorithm is:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers;
based on the cross-combination relationship between the features, for each feature xiTraining a vector v of size kiFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x;
adding the x vectors of Embedding of the characteristics of the users to obtain an Embedding vector representing the characteristics of the users, which is called User Embedding, measuring the distance between two User Embedding by using cosine distance, and if the distance between the two User Embedding is closer, indicating that the two users are more similar.
Preferably, firstly, randomly selecting K points as initial cluster centers, and distributing each point to the nearest cluster center by repeatedly calculating the distance from each point to the cluster center and adjusting the cluster center, so as to obtain K clusters;
and (3) clustering User Embedding of all users by using cosine distances, dividing the users into K clusters, and performing aggregation statistical calculation according to the favorite applets of each User calculated in the step 3 to obtain the popular applets in the clusters as the default recommended applets of the clusters.
The cold start recommendation system for the television small program provided by the invention comprises the following components:
module M1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode;
module M2: preprocessing collected behavior data, including data cleaning, marking, merging and weighting;
module M3: performing machine learning training by using a factor decomposition machine FM algorithm according to the preprocessed data to obtain User data User Embedding, and judging the similarity of users according to the distance between the User Embedding;
module M4: based on the distance between the User embeddings, clustering the User embeddings by using a K-means clustering algorithm;
module M5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user.
Preferably, the behavior data using movie on demand includes: behaviors of watching, collecting and canceling video contents, and names, duration and types of videos;
the behavior data using the television applet includes: the behavior of using, collecting and canceling the small program, and the name, the using time length and the type of the small program;
and storing the behavior data using the film and television on demand and the behavior data using the television small program in a database.
Preferably, for behavior data requested by using movies and televisions, marking the data according to the operation behavior and watching duration indexes of a user to represent the interest of the movies and televisions, sampling and combining the data according to the interest degree of the user to the movies and televisions according to a preset proportion to form a training data set;
for the behavior data using the television applet, the behaviors of using, collecting and canceling the collection are respectively given weight, and the favorite applet of each user is calculated.
Preferably, the formula of the FM algorithm is:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers;
based on the cross-combination relationship between the features, for each feature xiTraining a vector v of size kiFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x;
adding the x vectors of Embedding of the characteristics of the users to obtain an Embedding vector representing the characteristics of the users, which is called User Embedding, measuring the distance between two User Embedding by using cosine distance, and if the distance between the two User Embedding is closer, indicating that the two users are more similar.
Preferably, firstly, randomly selecting K points as initial cluster centers, and distributing each point to the nearest cluster center by repeatedly calculating the distance from each point to the cluster center and adjusting the cluster center, so as to obtain K clusters;
and (3) clustering User Embedding of all users by using cosine distances, dividing the users into K clusters, and performing aggregation statistical calculation according to the favorite applets of each User calculated in the step 3 to obtain the popular applets in the clusters as the default recommended applets of the clusters.
Compared with the prior art, the invention has the following beneficial effects:
according to the invention, embedding is trained through the film watching behaviors of the users, then clustering is carried out, the users are divided into a plurality of classes with similar interests, and then popular applets in each class are recommended to other users belonging to the class but not using the television applets for cold start.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
Example (b):
the invention provides a cold start recommendation scheme for a television applet, which trains embedding through the film watching behavior of a user, classifies the embedding of the user into a plurality of classes through a clustering algorithm, and the users in the same class have certain similarity in family structure and interest, extracts the favorite applet of the class of users and recommends the favorite applet to other users belonging to the class, and can be used as a cold start method for recommending the television applet.
As shown in fig. 1, the method specifically comprises the following steps:
step 1: the method comprises the steps of collecting behavior data of a user using video on demand through data dotting at a television end, wherein the behavior data comprises behaviors of watching, collecting and canceling video content and information of video name, duration, type and the like, and the data are stored in a database.
Step 2: behavior data of using the television small program by a user, such as behaviors of using, collecting and canceling the small program, information of name, using time length, type and the like of the small program, is collected by data dotting at the television end, and the data is stored in a database.
And step 3: and (3) cleaning the data collected in the steps (1) and (2), removing error data and extreme data, and processing the data. For the data collected in the step 1, marking marks on the data according to indexes such as operation behaviors of users, watching duration and the like, wherein the marks are used for showing the interest of the users in the movies, and the mark 1 is a positive example mark for showing the comparative interest of the users in the movie and television contents; 0 is a negative example flag indicating that the user is not so interested in the movie content; then the positive example and the negative example are respectively sampled and combined in proportion to form a training data set. And (3) giving certain weight to the behaviors of using, collecting, canceling, and the like of the data collected in the step (2) respectively, and calculating the favorite applet of each user.
And 4, step 4: and (3) performing machine learning training on the training data set obtained in the step (3), wherein the FM algorithm is used for performing data training. The FM (Factor Machine) algorithm is a Machine learning algorithm based on matrix decomposition, and is more suitable for solving the problem of feature combination in a large-scale sparse matrix. The formula for FM is as follows:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers;
based on the cross-combination relationship between the features, for each feature xiTraining a vector v of size kiFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x;
the Embedding of the features of the User is added to obtain the Embedding of the User, which is called User Embedding. The distance between two User embeddings can be measured as cosine distance, and if the distance between two User embeddings is closer, it indicates that the two users are more similar.
And 5: based on the distance of User Embedding, similar users can form a cluster, and users in the same cluster have certain commonality in family structure and hobbies. The K-Means clustering algorithm is used for clustering User Embedding, and the K-Means clustering method is a common clustering method. In the embodiment, cosine distances are used, User Embedding of all users is clustered, the users are divided into K clusters, aggregation statistical calculation is carried out according to the favorite applets of each User calculated in the step 3, hot applets in the clusters are obtained, and the hot applets can be used as default recommended applets of the clusters.
Step 6: since the User's smaller Embedding distance has greater similarity, even if a User (assumed to be User a) has not used any tv applet, it may prefer the applet that is preferred by the User similar to the User. Therefore, for a cold-start user A who does not use the television applet, as long as the user A has a watching behavior, the cold-start recommendation can be made, and the scheme is as follows:
and if the N users with the User Embedding distance nearest to the User A have favorite applets, recommending the applets to the User A, otherwise, recommending the popular applets in the cluster where the User A is located, which are calculated in the step 5, to the User A, so as to finish the cold start of the television applets of the User A.
The cold start recommendation system for the television small program provided by the invention comprises the following components: module M1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode; module M2: preprocessing collected behavior data, including data cleaning, marking, merging and weighting; module M3: performing machine learning training by using a factor decomposition machine FM algorithm according to the preprocessed data to obtain User data User Embedding, and judging the similarity of users according to the distance between the User Embedding; module M4: based on the distance between the User embeddings, clustering the User embeddings by using a K-means clustering algorithm; module M5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user.
The behavior data using movie and television on demand includes: behaviors of watching, collecting and canceling video contents, and names, duration and types of videos; the behavior data using the television applet includes: the behavior of using, collecting and canceling the small program, and the name, the using time length and the type of the small program; and storing the behavior data using the film and television on demand and the behavior data using the television small program in a database. Marking the data according to the operation behavior and watching duration indexes of the user for behavior data on demand by using the film and television to express the interest of the film and television, sampling and combining according to the interest degree of the user to the film and television according to a preset proportion to form a training data set; for the behavior data using the television applet, the behaviors of using, collecting and canceling the collection are respectively given weight, and the favorite applet of each user is calculated. The formula of the FM algorithm is:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers; based on the cross-combination relationship between the features, for each feature xiTraining a vector v of size kiFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x; adding the x vectors of Embedding of the features of the users to obtain an Embedding vector which can represent the features of the users and is called User Embedding, measuring the distance between two User Embedding by using cosine distance, and if the distance between two User Embedding is closer, indicating that the two users are more similar. Firstly, randomly selecting K points as initial cluster centers, and distributing each point to the nearest cluster center by repeatedly calculating the distance from each point to the cluster center and adjusting the cluster center to obtain K clusters; and (3) clustering User Embedding of all users by using cosine distances, dividing the users into K clusters, and performing aggregation statistical calculation according to the favorite applets of each User calculated in the step 3 to obtain the popular applets in the clusters as the default recommended applets of the clusters.
Those skilled in the art will appreciate that, in addition to implementing the systems, apparatus, and various modules thereof provided by the present invention in purely computer readable program code, the same procedures can be implemented entirely by logically programming method steps such that the systems, apparatus, and various modules thereof are provided in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Therefore, the system, the device and the modules thereof provided by the present invention can be considered as a hardware component, and the modules included in the system, the device and the modules thereof for implementing various programs can also be considered as structures in the hardware component; modules for performing various functions may also be considered to be both software programs for performing the methods and structures within hardware components.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention. The embodiments and features of the embodiments of the present application may be combined with each other arbitrarily without conflict.
Claims (10)
1. A method for recommending cold start of a television applet, comprising:
step 1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode;
step 2: preprocessing collected behavior data, including data cleaning, marking, merging and weighting;
and step 3: performing machine learning training by using a factor decomposition machine FM algorithm according to the preprocessed data to obtain User data User Embedding, and judging the similarity of users according to the distance between the User Embedding;
and 4, step 4: based on the distance between the User embeddings, clustering the User embeddings by using a K-means clustering algorithm;
and 5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user.
2. The method of claim 1, wherein the behavior data using video on demand comprises: behaviors of watching, collecting and canceling video contents, and names, duration and types of videos;
the behavior data using the television applet includes: the behavior of using, collecting and canceling the small program, and the name, the using time length and the type of the small program;
and storing the behavior data using the film and television on demand and the behavior data using the television small program in a database.
3. The method for recommending a cold start of a television applet as claimed in claim 1, wherein for the action data on demand using movies and televisions, the data are marked according to the operation action and watching duration index of the user to represent the interest of the movies and televisions, and are sampled and combined according to the interest degree of the user in the movies and televisions in a preset proportion to form a training data set;
for the behavior data using the television applet, the behaviors of using, collecting and canceling the collection are respectively given weight, and the favorite applet of each user is calculated.
4. The method of claim 1, wherein the FM algorithm has the formula:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers;
based on the cross-combination relationship between the features, for each feature xiTraining a k-sizeVector viFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x;
adding the x vectors of Embedding of the characteristics of the users to obtain an Embedding vector representing the characteristics of the users, which is called User Embedding, measuring the distance between two User Embedding by using cosine distance, and if the distance between the two User Embedding is closer, indicating that the two users are more similar.
5. The method of claim 3, wherein K clusters are obtained by randomly selecting K points as initial cluster centers, repeatedly calculating the distance from each point to the cluster center, and adjusting the cluster center to assign each point to the nearest cluster center;
and (3) clustering User Embedding of all users by using cosine distances, dividing the users into K clusters, and performing aggregation statistical calculation according to the favorite applets of each User calculated in the step 3 to obtain the popular applets in the clusters as the default recommended applets of the clusters.
6. A system for cold-start recommendation of a television applet, comprising:
module M1: collecting behavior data of a user using a film and television on demand and behavior data of a television applet at a television end in a data dotting mode;
module M2: preprocessing collected behavior data, including data cleaning, marking, merging and weighting;
module M3: performing machine learning training by using a factor decomposition machine FM algorithm according to the preprocessed data to obtain User data User Embedding, and judging the similarity of users according to the distance between the User Embedding;
module M4: based on the distance between the User embeddings, clustering the User embeddings by using a K-means clustering algorithm;
module M5: if the favorite applet of the User exists in the N users which are closest to the User Embedding distance of the User, recommending the favorite applet to the User; otherwise, calculating and recommending the hot small programs in the cluster where the user is located, so as to finish the cold start of the television small programs of the user.
7. The system of claim 6, wherein the behavioral data for video on demand includes: behaviors of watching, collecting and canceling video contents, and names, duration and types of videos;
the behavior data using the television applet includes: the behavior of using, collecting and canceling the small program, and the name, the using time length and the type of the small program;
and storing the behavior data using the film and television on demand and the behavior data using the television small program in a database.
8. The system of claim 6, wherein for the action data requested by the movie, the data is marked according to the operation action and watching duration index of the user to show the interest of the movie, and the data is sampled and combined according to the interest degree of the movie and the television in a preset proportion to form a training data set;
for the behavior data using the television applet, the behaviors of using, collecting and canceling the collection are respectively given weight, and the favorite applet of each user is calculated.
9. The system of claim 6, wherein the formula of the FM algorithm is:
wherein the content of the first and second substances,
is a conventional logistic regression model, w0Is the initial weight, wiRepresents each feature xiCorresponding weight value, n is total number of features, i and j are serial numbers;
based on the cross-combination relationship between the features, for each feature xiTraining a vector v of size kiFeature xixjCombined weight value by vector viAnd vjInner product of (2)<vi,vj>Representing, namely training a vector v obtained through an FM algorithm to be an embedding expression of the feature x;
adding the x vectors of Embedding of the characteristics of the users to obtain an Embedding vector representing the characteristics of the users, which is called User Embedding, measuring the distance between two User Embedding by using cosine distance, and if the distance between the two User Embedding is closer, indicating that the two users are more similar.
10. The system of claim 8, wherein K points are randomly selected as initial cluster centers, and the K clusters are obtained by repeatedly calculating the distance between each point and the cluster center and adjusting the cluster center to assign each point to the nearest cluster center;
and (3) clustering User Embedding of all users by using cosine distances, dividing the users into K clusters, and performing aggregation statistical calculation according to the favorite applets of each User calculated in the step 3 to obtain the popular applets in the clusters as the default recommended applets of the clusters.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111356474.6A CN114268836A (en) | 2021-11-16 | 2021-11-16 | Cold start recommendation method and system for television applet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111356474.6A CN114268836A (en) | 2021-11-16 | 2021-11-16 | Cold start recommendation method and system for television applet |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114268836A true CN114268836A (en) | 2022-04-01 |
Family
ID=80825101
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111356474.6A Pending CN114268836A (en) | 2021-11-16 | 2021-11-16 | Cold start recommendation method and system for television applet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114268836A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110225A (en) * | 2019-04-17 | 2019-08-09 | 重庆第二师范学院 | Online education recommended models and construction method based on user behavior data analysis |
| CN112528164A (en) * | 2020-12-14 | 2021-03-19 | 建信金融科技有限责任公司 | User collaborative filtering recall method and device |
| CN112650940A (en) * | 2019-10-10 | 2021-04-13 | 北京多点在线科技有限公司 | Recommendation method and device of application program, computer equipment and storage medium |
-
2021
- 2021-11-16 CN CN202111356474.6A patent/CN114268836A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110225A (en) * | 2019-04-17 | 2019-08-09 | 重庆第二师范学院 | Online education recommended models and construction method based on user behavior data analysis |
| CN112650940A (en) * | 2019-10-10 | 2021-04-13 | 北京多点在线科技有限公司 | Recommendation method and device of application program, computer equipment and storage medium |
| CN112528164A (en) * | 2020-12-14 | 2021-03-19 | 建信金融科技有限责任公司 | User collaborative filtering recall method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108009228B (en) | Method and device for setting content label and storage medium | |
| CN109189951B (en) | Multimedia resource recommendation method, equipment and storage medium | |
| US10277696B2 (en) | Method and system for processing data used by creative users to create media content | |
| CN112395489B (en) | Recommendation method, recommendation device, recommendation equipment and computer storage medium | |
| CN107888950A (en) | A kind of method and system for recommending video | |
| US20100070507A1 (en) | Hybrid content recommending server, system, and method | |
| US20120323725A1 (en) | Systems and methods for supplementing content-based attributes with collaborative rating attributes for recommending or filtering items | |
| CN102523511A (en) | Network program aggregation and recommendation system and network program aggregation and recommendation method | |
| CN108108491B (en) | Multimedia data recommendation method and device | |
| CN104182449A (en) | System and method for personalized video recommendation based on user interests modeling | |
| CN108351897B (en) | Methods, systems, and media for creating and updating groups of media content items | |
| CN104199896A (en) | Video similarity determining method and video recommendation method based on feature classification | |
| CN111861550B (en) | Family portrait construction method and system based on OTT equipment | |
| WO2023040506A1 (en) | Model-based data processing method and apparatus, electronic device, computer-readable storage medium, and computer program product | |
| CN109753601A (en) | Recommendation information clicking rate determines method, apparatus and electronic equipment | |
| WO2023231542A1 (en) | Representation information determination method and apparatus, and device and storage medium | |
| US20140172501A1 (en) | System Apparatus Circuit Method and Associated Computer Executable Code for Hybrid Content Recommendation | |
| CN102572543A (en) | Digital television program recommending system and method thereof | |
| CN112825089A (en) | Article recommendation method, article recommendation device, article recommendation equipment and storage medium | |
| CN112199582A (en) | Content recommendation method, device, equipment and medium | |
| CN112579822A (en) | Video data pushing method and device, computer equipment and storage medium | |
| CN109063080B (en) | Video recommendation method and device | |
| CN113327132A (en) | Multimedia recommendation method, device, equipment and storage medium | |
| CN110188277B (en) | Resource recommendation method and device | |
| CN112733006B (en) | User portrait generation method, device and 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 |