CN109587527B - Personalized video recommendation method and device - Google Patents
Personalized video recommendation method and device Download PDFInfo
- Publication number
- CN109587527B CN109587527B CN201811330192.7A CN201811330192A CN109587527B CN 109587527 B CN109587527 B CN 109587527B CN 201811330192 A CN201811330192 A CN 201811330192A CN 109587527 B CN109587527 B CN 109587527B
- Authority
- CN
- China
- Prior art keywords
- video content
- user
- item
- preference value
- video
- 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.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/251—Learning process for intelligent management, e.g. learning user preferences for recommending movies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/258—Client or end-user data management, e.g. managing client capabilities, user preferences or demographics, processing of multiple end-users preferences to derive collaborative data
- H04N21/25866—Management of end-user data
- H04N21/25891—Management of end-user data being end-user preferences
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/43—Processing of content or additional data, e.g. demultiplexing additional data from a digital video stream; Elementary client operations, e.g. monitoring of home network or synchronising decoder's clock; Client middleware
- H04N21/442—Monitoring of processes or resources, e.g. detecting the failure of a recording device, monitoring the downstream bandwidth, the number of times a movie has been viewed, the storage space available from the internal hard disk
- H04N21/44213—Monitoring of end-user related data
- H04N21/44222—Analytics of user selections, e.g. selection of programs or purchase activity
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4667—Processing of monitored end-user data, e.g. trend analysis based on the log file of viewer selections
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/466—Learning process for intelligent management, e.g. learning user preferences for recommending movies
- H04N21/4668—Learning process for intelligent management, e.g. learning user preferences for recommending movies for recommending content, e.g. movies
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Social Psychology (AREA)
- Computer Networks & Wireless Communication (AREA)
- Computing Systems (AREA)
- Computer Graphics (AREA)
- Two-Way Televisions, Distribution Of Moving Picture Or The Like (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a method and a device for recommending personalized videos, and solves the problem that a video platform cannot accurately recommend video contents with personalized preferences for users. The method comprises the following steps: obtaining an association item related to any video content, determining a preference value of each associated video content of the association item according to a video browsing history record, normalizing the preference value of each associated video content to obtain a normalized preference value of each associated video content, obtaining the click probability of each associated video content of the association item by the user according to the normalized preference value of each associated video content by using a random walk recommendation algorithm, and pushing the preference video content under the association item to the user according to the click probability.
Description
Technical Field
The invention relates to the technical field of content recommendation, in particular to a personalized video recommendation method and device.
Background
With the development of scientific information technology, for content-based recommendations, sufficient accuracy of media tags is required, and for video recommendations for users, many video platforms are currently directed to video recordings that users have viewed on the platform, recommending video content related to the video content watched by the user to the user according to the click amount or comment of the user on the video and the like, or recommending more popular video content under the video platform to the user according to some video tags or associated tags, in the video recommendation algorithm, however, the video tag cannot accurately reflect the preference degree of the user for the recommended video content, and the video platform can not accurately know the preference degree of the user to the recommended video content due to the fact that the user can not click or comment the video, and therefore the video platform can not accurately recommend the video content with personalized preference for the user.
Disclosure of Invention
The invention provides a method and a device for personalized video recommendation based on weighted random walk, and solves the problem that a video platform cannot accurately recommend video content with personalized preferences for a user.
According to a first aspect of the invention, there is provided a method of personalizing video recommendations, the method comprising:
acquiring related items related to any video content;
determining a preference value of each associated video content of the associated item according to the video browsing history;
normalizing the preference value of each associated video content to obtain a normalized preference value of each associated video content;
obtaining the click probability of the user on each associated video content of the associated item by using a random walk recommendation algorithm according to the normalized preference value of each associated video content;
and pushing the preference video content under the associated item to the user according to the click probability.
In accordance with a second aspect of the present invention, there is provided an apparatus for personalizing video recommendations, the apparatus comprising a processor and a memory, wherein:
the memory is used for storing an executable program;
the processor, when executing the executable program, implements the method of any of claims 1-7.
Compared with the prior art, the personalized recommendation method provided by the invention has the following beneficial effects:
in the method, a personalized video recommendation method based on the combination of the weight of the recommendation-associated content item and the random walk recommendation algorithm is provided, so that the personalized recommendation can be performed on a certain content recommendation item independently, the weights of a plurality of content recommendation items can be fused, the recommendation is combined, the accuracy of the personalized recommendation is improved, and the like.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
Fig. 1 is a schematic diagram of a personalized video recommendation method according to a first embodiment;
FIG. 2 is a diagram illustrating a user behavior data set according to one embodiment;
fig. 3 is a schematic device diagram of a personalized video recommendation device according to a second embodiment;
fig. 4 is a schematic device structure diagram of a personalized video recommendation device according to a fourth embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail with reference to the accompanying drawings, and it is apparent that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Hereinafter, some terms in the embodiments of the present application are explained to facilitate understanding by those skilled in the art.
(1) In the embodiments of the present application, the term "plurality" means two or more, and other terms are similar thereto.
(2) "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
The first embodiment is as follows:
the invention provides a personalized video recommendation method based on weighted random walk, a schematic diagram of which is shown in figure 1, and the method specifically comprises the following steps:
optionally, the association item may be obtained from a tag related to the video content in the video library, the association item may be, but is not limited to, a video name/actor/director, and a person skilled in the art may set other association items that may be used to recommend the video content to the user according to actual needs;
each of the associated items includes at least one associated video content, for example, when the associated item is a video name, the associated video content may be a movie in arylua or a tv show in Rezhu Gege; when the associated item is an actor, the associated video content can be Huangxuan, fanbing ice and the like; when the above-mentioned related item is director, the above-mentioned related video content can be Zhangyi collusion, yellow dawn and so on;
optionally, in implementation, the association may be input by a user, or preset by a person skilled in the art.
optionally, in a specific implementation, according to a video browsing history of a user, obtaining an actual viewing duration and a number of clicks of each associated video content by the user, or an index that can reflect a degree of preference of the user for different video contents, such as a bullet screen generated for each associated video content, and determining a preference value of each associated video content of the associated item according to the index that reflects the degree of preference of the user for different video contents;
as an optional manner, in this embodiment, the actual viewing duration of each associated video content by the user reflects the preference degree of the user for different video contents, which is specifically implemented as follows:
determining the total time length of each associated video content of the associated item, determining the watching time length of each associated video content of the associated item by the user according to the video browsing history, and dividing the watching time length of each associated video content by the total time length of each associated video content to obtain the preference of the user for each associated video content;
optionally, when determining the viewing duration of each associated video content of the associated item by the user, performing a deduplication operation on the viewing duration, that is, acquiring all video browsing history records of each associated video content of the associated item by the user, and determining all viewing duration records of each associated video content by the user according to the video browsing history records;
and deleting the repeated time length corresponding to the repetition of the video content from all the watching time length records to obtain the watching time length of each associated video content.
optionally, in implementation, the preference value of each associated video content may be normalized according to a preset rule, and the proportion of the preference value of each associated video content is determined by combining the preference values of all associated video contents in the associated item by the user;
as an implementable manner, the preset rule may be:
summing the preference values of all the associated video contents of the associated items to obtain a sum of preference values, and dividing the preference value of each associated video content by the sum of preference values to obtain a normalized preference value of each associated video content;
the above normalized preference value riThe calculation formula of (a) is as follows:
in the above formula, vkA preference value for each associated video content for the user.
in a specific implementation, according to the obtained normalized preference value of each associated video content and a webpage link corresponding to each associated video content, a webpage ranking algorithm is combined, and a random walk recommendation algorithm is utilized to obtain the click probability of each associated video content webpage;
the web page ranking algorithm may be a web page ranking algorithm pagerank, which is an algorithm for measuring the importance of a specific web page relative to other web pages in a search engine, and the calculation result is used as an important index of the web page ranking in the search result of the search engine google. The web pages are connected with each other through hyperlinks, and countless web pages on the Internet form an oversized graph. The pagerank assumes that a user randomly selects a webpage from all webpages to browse, and then directly skips the webpage without stopping through the hyperlink. After reaching each web page, the user has two choices: when the user finishes browsing or continues to rotate a link for browsing, the algorithm presets that the probability of the user continuing to browse is d, the user randomly selects one of all hyperlinks in the current page with equal probability to continue browsing, which is a random walk process, after many times of such walks, the probability of each webpage visited by the user converges to a stable value, the probability is an importance index of the webpage and is used for webpage ranking, and the algorithm iteration relationship is shown as the following formula 1:
equation 1:
PR (i) and PR (j) in the formula 1 are pagerank values of the web pages i and j respectively, the pagerank value is a feature vector in a special matrix and represents the ranking importance of a certain web page, and the higher the PR value is, the more important/popular the connection is;
as with equation 1 above, the calculation of PR (pagerank) for web page i is based on the following two basic assumptions:
the quantity assumes that in the Web graph model, if the number of the inbound links pointed by other Web pages received by a page node is more, the more important the page is;
the quality assumption is that: the incoming links to page i are of different quality, and the high quality page may pass more weight to other pages through the links. Therefore, the more the high-quality page points to the page i, the more important the page i is;
i and j in the formula 1 are labels of web pages, N in the formula is the number of all web pages, and j in the formula belongs to in (i) and represents all web pages entering the web page i; out (j) represents the total number of pages linked out of page j (the PR value of page j for each page is divided equally according to the number of out-links of page j);
d in the above formula 1 is a damping coefficient, and d has a meaning that at any time, the probability that the user will continue to browse the next webpage backwards after reaching a certain page, the user stops clicking at the probability of (1-d) and randomly jumps to any webpage, the initial value of d can be set by a person skilled in the art according to the actual situation, and the initial value of d is generally set to 0.85;
D in the above formula 1 is a damping coefficient, and d has a meaning that at any time, the probability that the user will continue to browse the next webpage backwards after reaching a certain page, the user stops clicking at the probability of (1-d) and randomly jumps to any webpage, the initial value of d can be set by a person skilled in the art according to the actual situation, and the initial value of d is generally set to 0.85; based on the random walk algorithm, the invention combines the preference value of the user to the video content to carry out random walk of the following formula 2 on all the webpages related to the video content:
equation 2:
in the above formula 2, riFor the normalized preference value of the certain associated video content item, i in the above formula 2 is a tag of a web page link corresponding to the certain associated video content item, and pr (i) is a preference value of the user for the associated video content of the associated item corresponding to the web page link i; PR (j) has the same meaning as that expressed by the above formula 1, wherein j ∈ in (i) in the formula represents all webpage links entering the webpage i; l e out (j) represents all web page links out of web page j link; the meaning of d in the above formula 2 is the same as that of the above formula 1;
sigma in the above formulal∈out(j)rlRepresenting the sum of all web page links linked out of web page link j.
In this embodiment, the PR value of the web page link of each related video content in the related item, which is calculated according to the above method, is used as the click probability of the user for each related video content in the related item.
And 150, pushing the preference video content under the associated item to the user according to the click probability.
Optionally, in implementation, the user may rank the click probability of each associated video content from high to low, and push any one or any plurality of video contents ranked the top to the user according to different scenes.
By using the method provided by the invention, the preference value of each associated video content of any associated item, which is obtained based on the record of watching videos of the user, is used as the weight, the iterative relationship of the pagerank algorithm is improved, the PR value of the webpage link corresponding to each associated video content of any associated item is calculated, the PR value is used as the click probability of the user to each associated video content of the associated item, and when the video content preferred by the user is recommended to the user according to different click probabilities, the video content preferred by the user can be more accurately obtained, and the accuracy of personalized recommendation is improved.
When the association item is a main association item related to the video content and a normalized preference value of each associated video content is obtained, the method further comprises the following steps:
determining a preference value of at least one auxiliary associated video content of at least one auxiliary associated item associated with the main associated item according to the video browsing history;
determining a correlation coefficient corresponding to each auxiliary correlation item, multiplying and summing the correlation coefficient with the preference value of the at least one corresponding auxiliary correlation video content, summing the result with the preference value of each correlation video content in the main correlation item, and updating the summation result into the preference value of each correlation video content of the main correlation item;
optionally, determining a correlation coefficient corresponding to each auxiliary correlation item according to the clicked amount of the correlated video content under the correlation item, in implementation, a person skilled in the art may also determine a correlation coefficient corresponding to each auxiliary correlation item according to other factors affecting the importance of the correlation item;
as an optional mode, according to the video browsing history of all users, determining the association coefficient corresponding to each auxiliary association item through a mode of step length 0.1 and an on-line ab _ test system test.
A specific process of recommending the video content to the user 1 is given below:
step 1101, acquiring any video content related association item, and entering step 1201;
in this embodiment, it is assumed that the associated item related to any video content is obtained as a video name, and the associated content items of the video name are "arylwa", "tuina", and "my father mother".
Step 1201, determining a preference value of each associated video content of the associated item according to the video browsing history record, and entering step 1301;
in this embodiment, assuming that the user 1 browses the first 20 minutes of the movie in the form of "arylghua" for the first time, browses the lower 10-30 minutes of the movie in the form of "arylghua" for the second time, and browses the last 18 minutes of the movie in the form of "arylghua" for the third time, the obtained watching time of the user on the associated video content in the form of "arylghua" is 60 minutes, and the total time of the "arylghua" is 120 minutes, it is determined that the preference value of the user 1 on the "arylghua" is ((30+10+18)/120), which is 0.4;
based on the method, the preference values of the user 1 to tuina and my father and mother are respectively determined;
when the association item is determined to be a video name, and an auxiliary association item actor and an auxiliary association video content actor yellow pavilion and actor seedlings are provided, when the preference value of the user 1 to the fanghua is calculated, if the preference value of the user 1 to the actor yellow pavilion in the fanghua is calculated to be 0.5 value, the preference value of the actor seedlings is 0.3 and the association coefficient of the auxiliary association item is 0.4 according to the method of the step 102, the preference value of the user 1 to the fanghua is calculated to be 0.4+ (0.3+0.5) × 0.4, namely 0.64;
if the determined auxiliary associated video content is actor yellow pavilion and director artistic collusion when the main associated item is a video name and the determined auxiliary associated item is actor and director, and if the preference value of the user 1 for the actor yellow pavilion is calculated according to the method, the preference value of the user 1 for the actor yellow pavilion is calculated to be 0.5, the preference value of the user 1 for the director artistic collusion is calculated to be 0.6, the association coefficient of the actor of the auxiliary associated item is 0.4, and the association coefficient of the director is 0.2, the preference value of the user 1 for the actor yellow pavilion is calculated to be 0.4+0.5 + 0.4+0.6+ 0.2, namely 0.72; calculating the preference value of the user 1 to tuina as 0.6+0.5 x 0.4, namely 0.8; calculating the preference value of the user 1 to My father mother as 0.8+0.6 x 0.2, namely 0.92; the director of the aforementioned fanghua and my father mother is Zhang artistic conspiracy, and the aforementioned actor Huangxuan has played the role of fanghua and tuina.
Step 1301, normalizing the preference value of each associated video content to obtain a normalized preference value of each associated video content, and entering step 1401;
optionally, before normalizing the preference value of each of the associated video contents, a user behavior data set of the user and all associated video contents under the associated item may be established, where the user behavior data set may be composed of individual binary groups (u, m), as shown in fig. 2, where u represents different users, m represents different associated video contents, and data on different sides between u and m represents preference values of the user for different associated video contents; the user behavior data set may also be a relational list, as shown in table 1 below, where u represents different users, m represents different associated video content, and data in the table that differs between u and m represents user preference values for different associated video content.
| m1 | m2 | m3 | |
| u1 | 0.6 | 0.4 | |
| u2 | 0.2 | 0.6 |
TABLE 1
The specific form of the user behavior data set is not limited too much, and the user behavior data set can be set by a person skilled in the art according to the actual situation;
in this embodiment, according to the above content, a user behavior data set of all associated video contents of the user and the above video name is established, as shown in table 2 below:
| fanghua tea | Massage and manipulation | My | |
| User | |||
| 1 | 0.4 | 0.6 | 0.8 |
| User 2 | 0.4 | 0.6 |
TABLE 2
According to the above table 2, the preference values of all the associated content items under the video name of the above user 1 are normalized as follows:
and determining the normalized preference value of the user 1 to the movie's ' arylghua ' as a value (0.4/(0.4+0.6+0.8)), namely 0.22, and respectively determining the normalized preference values of the user 1 to ' tuina ' and ' my father mother ' according to a method for determining the normalized preference value of the user 1 to the ' arylghua '.
1401, obtaining the click probability of the user on each associated video content of the association item by using a random walk recommendation algorithm according to the normalized preference value of each associated video content, and entering 1501;
from the above step 1201, if it is calculated that the preference values of "arylua", "tuina", and "my father mother" of the user 1 are 0.72, 0.8, and 0.92, respectively, the normalization calculation is performed on 0.72, 0.8, and 0.92, and the normalized values of 0.30, 0.33, and 0.38 are used as r of the above formula 2, respectivelyiThe values of (2) are calculated, and the PR values of the web page links of my father mother, arylwa, and tuina, which are associated with the video names, are obtained by using the above formula 2, respectively, as the click probabilities of the user 1 on arylwa, tuina, and my father mother.
Step 1501, pushing the preference video content under the relevant item to the user according to the click probability.
Optionally, in implementation, the user may rank the click probability of each associated video content from high to low, and push any one or any plurality of video contents ranked the top to the user.
In the order of the click probability of the user 1 on the aryls, the tuina and the my father mother, which is obtained in the step 1401, from high to low, any one or any plurality of video contents with the top ranking are pushed to the user.
Example two:
the embodiment is a personalized video recommendation device based on weighted random walk, and referring to fig. 3, the device includes a processor 300 and a memory 301, and the principle of the device to solve the problem is similar to that of the method, so the implementation of the device may refer to the implementation of the method, and repeated details are not repeated.
Wherein:
the memory is used for storing an executable program;
the processor is configured to implement the method according to the first embodiment when executing the executable program.
Optionally, the processor is specifically configured to:
acquiring related items related to any video content; the associated item of any of the above video contents is a video title/actor/director.
Determining a preference value of each associated video content of the associated items according to the video browsing history;
normalizing the preference value of each associated video content to obtain a normalized preference value of each associated video content;
obtaining the click probability of the user on each associated video content of the associated item by using a random walk recommendation algorithm according to the normalized preference value of each associated video content;
and pushing the preference video content under the associated item to the user according to the click probability.
The processor is further configured to, after determining the preference value for each associated video content of the associated item:
determining a preference value of at least one auxiliary associated video content of at least one auxiliary associated item associated with the main associated item according to the video browsing history;
determining a correlation coefficient corresponding to each auxiliary correlation item, multiplying and summing the correlation coefficient with the preference value of the at least one corresponding auxiliary correlation video content, summing the result with the preference value of each correlation video content in the main correlation item, and updating the summation result into the preference value of each correlation video content of the main correlation item;
the processor is specifically configured to determine a total duration of each associated video content of the associated item, determine, according to the video browsing history, a viewing duration of each associated video content of the associated item by the user, and divide the viewing duration of each associated video content by the total duration of each associated video content to obtain a preference value of the user for each associated video content.
The processor is specifically configured to determine all viewing duration records of the user for each associated video content according to the video browsing history record, and delete a repetition duration corresponding to the repetition of the video content from all the viewing duration records to obtain the viewing duration of each associated video content.
The processor is specifically configured to sum preference values of all associated video contents of the associated item to obtain a sum of preference values, and divide the preference value of each associated video content by the sum of preference values to obtain a normalized preference value of each associated video content.
The processor is specifically configured to rank the click probability of each associated video content from high to low for the user, and push any one or any plurality of video contents ranked first to the user.
The processor is specifically configured to, optionally, determine a relevance coefficient corresponding to each auxiliary relevance item according to a clicked amount of the relevant video content under the relevance item.
Example three:
the present embodiment is a computer storage medium storing a computer program that realizes the contents of any one of the first and second embodiments when executed.
Example four:
the embodiment provides a personalized video recommendation method and device based on weighted random walk, a schematic structural diagram of the device is shown in fig. 4, and the device includes:
an associated information acquiring unit 401, configured to acquire an associated item related to any video content;
a preference value calculating unit 402, configured to determine a preference value of each associated video content of the associated item according to the video browsing history;
a normalizing unit 403, configured to normalize the preference value of each related video content to obtain a normalized preference value of each related video content;
a video recommending unit 404, configured to obtain, according to the normalized preference value of each associated video content, a click probability of the user on each associated video content of the associated item by using a random walk recommendation algorithm; and pushing the preference video content under the associated item to the user according to the click probability.
The associated item of any of the above video contents is a video title/actor/director.
The preference value calculating unit is further configured to, after the association is a main association associated with the video content and a normalized preference value of each associated video content is obtained:
determining a preference value of at least one auxiliary associated video content of at least one auxiliary associated item associated with the main associated item according to the video browsing history;
and determining the association coefficient corresponding to each auxiliary association item, multiplying and summing the association coefficient with the preference value of the at least one corresponding auxiliary association video content, and summing the result of the preference value of each associated video content in the main association item to obtain the preference value of each associated video content of the main association item.
The preference value calculating unit is specifically configured to determine a total duration of each associated video content of the associated item, determine a viewing duration of each associated video content of the associated item by the user according to a video browsing history, and divide the viewing duration of each associated video content by the total duration of each associated video content to obtain a preference value of the user for each associated video content.
The preference value calculating unit is specifically configured to determine all viewing duration records of the user for each associated video content according to the video browsing history record, and delete a duration corresponding to the repetition of the video content from all the viewing duration records to obtain the viewing duration of each associated video content.
The normalization unit is specifically configured to sum preference values of all associated video contents of the associated item to obtain a sum of preference values, and divide the preference value of each associated video content by the sum of preference values to obtain a normalized preference value of each associated video content.
The video recommending unit is specifically configured to rank the click probability of each associated video content from high to low for the user, and push any one or any plurality of video contents ranked first to the user.
The preference value calculating unit is specifically configured to determine a relevance coefficient corresponding to each auxiliary relevance item according to a clicked amount of the relevant video content under the relevance item.
It should be noted that the technical solutions of the embodiments of the present invention can be combined with each other, but must be based on the realization of the technical solutions by those skilled in the art, and when the technical solutions are contradictory or can not be realized, the combination of the technical solutions should be considered to be absent and not to be within the protection scope of the present invention. The above description is only for the preferred embodiment of the present invention and is not intended to limit the scope of the present invention, and all equivalent structural changes made by using the contents of the present specification and the drawings, or any other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A method for personalized video recommendation, comprising:
acquiring related items related to any video content;
determining a preference value of each associated video content of the associated item according to a video browsing history of a user;
normalizing the preference value of each associated video content to obtain a normalized preference value of each associated video content;
obtaining the click probability of the user on each associated video content of the associated item by using a random walk recommendation algorithm according to the normalized preference value of each associated video content;
pushing preference video content under the associated item to a user according to the click probability;
obtaining the click probability of each associated video content of the associated item by the user by using a random walk recommendation algorithm, wherein the method comprises the following steps:
will be according to the formula
Calculating a PR value of a webpage link of each associated video content of the associated item, wherein the PR value is used as the click probability of the user on each associated video content of the associated item;
wherein d is a damping coefficient, riThe normalized preference value of a certain associated video content, i is a label of a webpage link corresponding to the certain associated video content, j is a label of a webpage, PR (i) is a preference value of the user on the associated video content corresponding to the webpage link i of the associated item, PR (j) is a ranking importance pagerank value of the webpage j, the pagerank value is a feature vector in a matrix and represents the ranking importance of a certain webpage, and j belongs to in (i) represents all webpage links entering the webpage i; l e out (j) represents all web page links out of web page j link; sigmal∈out(j)rlRepresenting the sum of all web page links linked out of web page link j.
2. The method of claim 1, wherein determining the preference value for each associated video content of the associated item further comprises:
determining a preference value of at least one auxiliary associated video content of at least one auxiliary associated item associated with the main associated item according to the video browsing history;
and after multiplying and summing the correlation coefficient corresponding to each auxiliary correlation item with the preference value of the corresponding at least one auxiliary correlation video content, summing the results of the preference values of the correlation video contents in the main correlation item associated with the auxiliary correlation video contents, and updating the summation result to the preference value of each correlation video content of the main correlation item.
3. The method of claim 1, wherein determining a preference value for each associated video content of the associated item based on a video browsing history comprises:
determining a total duration of each associated video content of the associated item;
determining the watching time of the user on each associated video content of the associated item according to the video browsing history;
and dividing the watching duration of each associated video content by the total duration of each associated video content to obtain the preference value of the user for each associated video content.
4. The method of claim 3, wherein determining the viewing duration of each associated video content of the associated item by the user based on the video browsing history comprises:
determining all watching duration records of the user on each associated video content according to the video browsing history record;
and deleting the repeated time length corresponding to the repetition of the video content from all the watching time length records to obtain the watching time length of each associated video content.
5. The method of claim 1, wherein normalizing the preference value for each associated video content comprises:
summing preference values of all the associated video contents of the associated items to obtain a sum of the preference values;
and dividing the preference value of each associated video content by the sum of the preference values to obtain the normalized preference value of each associated video content.
6. The method of claim 1, wherein pushing the user with the preferred video content under the associated term according to the click probability comprises:
and sequencing the click probability of the user on each associated video content from high to low, and pushing any one or more video contents with the highest sequencing to the user.
7. The method of claim 2, wherein determining the correlation coefficient corresponding to each secondary correlation comprises:
and determining the association coefficient corresponding to each auxiliary association item according to the clicked amount of the auxiliary association video content under each auxiliary association item.
8. The method of claim 1, wherein the association comprises:
video title or actor or director.
9. A personalized video recommendation device is characterized in that the device
Comprising a processor and a memory, wherein:
the memory is used for storing an executable program;
the processor, when executing the executable program, implements the method of any of claims 1-8.
10. A computer storage medium, characterized in that the computer storage medium stores a computer program which, when executed by a processor, implements a method of personalized video recommendation according to any of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811330192.7A CN109587527B (en) | 2018-11-09 | 2018-11-09 | Personalized video recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811330192.7A CN109587527B (en) | 2018-11-09 | 2018-11-09 | Personalized video recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109587527A CN109587527A (en) | 2019-04-05 |
| CN109587527B true CN109587527B (en) | 2021-06-04 |
Family
ID=65921953
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811330192.7A Active CN109587527B (en) | 2018-11-09 | 2018-11-09 | Personalized video recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109587527B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110609856B (en) * | 2019-08-22 | 2023-06-13 | 贵州省广播电视信息网络股份有限公司 | Method for recommending AB report statistics based on artificial intelligence |
| CN112714359B (en) * | 2019-10-24 | 2022-09-09 | 北京达佳互联信息技术有限公司 | Video recommendation method and device, computer equipment and storage medium |
| CN111708901B (en) * | 2020-06-19 | 2023-10-13 | 腾讯科技（深圳）有限公司 | Multimedia resource recommendation method and device, electronic equipment and storage medium |
| CN116089732B (en) * | 2023-04-11 | 2023-07-04 | 江西时刻互动科技股份有限公司 | User preference identification method and system based on advertisement click data |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001318908A (en) * | 2000-04-07 | 2001-11-16 | Nec Corp | System and method for using random walk in order to search connection and use of web page |
| CN102799671A (en) * | 2012-07-17 | 2012-11-28 | 西安电子科技大学 | Network individual recommendation method based on PageRank algorithm |
| CN104935963A (en) * | 2015-05-29 | 2015-09-23 | 中国科学院信息工程研究所 | Video recommendation method based on timing sequence data mining |
| CN105159930A (en) * | 2015-08-05 | 2015-12-16 | 百度在线网络技术（北京）有限公司 | Search keyword pushing method and apparatus |
| CN106156113A (en) * | 2015-04-03 | 2016-11-23 | Tcl集团股份有限公司 | A kind of method of video recommendations, system and electronic equipment |
| CN106372101A (en) * | 2016-07-20 | 2017-02-01 | Tcl集团股份有限公司 | Video recommendation method and apparatus |
| CN107918652A (en) * | 2017-11-15 | 2018-04-17 | 浙江大学 | A kind of method that the film recommendation based on social networks is carried out using multi-modal e-learning |
| CN108197244A (en) * | 2017-12-29 | 2018-06-22 | 北京奇虎科技有限公司 | It is a kind of to search for the method for pushing and device for recommending word |
-
2018
- 2018-11-09 CN CN201811330192.7A patent/CN109587527B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001318908A (en) * | 2000-04-07 | 2001-11-16 | Nec Corp | System and method for using random walk in order to search connection and use of web page |
| CN102799671A (en) * | 2012-07-17 | 2012-11-28 | 西安电子科技大学 | Network individual recommendation method based on PageRank algorithm |
| CN106156113A (en) * | 2015-04-03 | 2016-11-23 | Tcl集团股份有限公司 | A kind of method of video recommendations, system and electronic equipment |
| CN104935963A (en) * | 2015-05-29 | 2015-09-23 | 中国科学院信息工程研究所 | Video recommendation method based on timing sequence data mining |
| CN105159930A (en) * | 2015-08-05 | 2015-12-16 | 百度在线网络技术（北京）有限公司 | Search keyword pushing method and apparatus |
| CN106372101A (en) * | 2016-07-20 | 2017-02-01 | Tcl集团股份有限公司 | Video recommendation method and apparatus |
| CN107918652A (en) * | 2017-11-15 | 2018-04-17 | 浙江大学 | A kind of method that the film recommendation based on social networks is carried out using multi-modal e-learning |
| CN108197244A (en) * | 2017-12-29 | 2018-06-22 | 北京奇虎科技有限公司 | It is a kind of to search for the method for pushing and device for recommending word |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109587527A (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110941740B (en) | Video recommendation method and computer-readable storage medium | |
| CN109587527B (en) | Personalized video recommendation method and device | |
| US9355168B1 (en) | Topic based user profiles | |
| CN109819284B (en) | Short video recommendation method and device, computer equipment and storage medium | |
| US10572565B2 (en) | User behavior models based on source domain | |
| US9978093B2 (en) | Method and system for pushing mobile application | |
| US8015192B2 (en) | Cliprank: ranking media content using their relationships with end users | |
| US9348935B2 (en) | Systems and methods for augmenting a keyword of a web page with video content | |
| US8370337B2 (en) | Ranking search results using click-based data | |
| US20190171725A1 (en) | Building user profiles by relevance feedback | |
| US9706008B2 (en) | Method and system for efficient matching of user profiles with audience segments | |
| US9098551B1 (en) | Method and system for ranking content by click count and other web popularity signals | |
| US20090006368A1 (en) | Automatic Video Recommendation | |
| US20140279793A1 (en) | Systems and methods for providing relevant pathways through linked information | |
| US20070219994A1 (en) | Methods and systems for displaying media utilizing user-generated data | |
| US20090077081A1 (en) | Attribute-Based Item Similarity Using Collaborative Filtering Techniques | |
| US20110145226A1 (en) | Product similarity measure | |
| US20150039608A1 (en) | Media content rankings for discovery of novel content | |
| US20120284283A1 (en) | Information Processing Method, Apparatus, and Computer Program | |
| US20170199930A1 (en) | Systems Methods Devices Circuits and Associated Computer Executable Code for Taste Profiling of Internet Users | |
| US9424353B2 (en) | Related entities | |
| US20150310529A1 (en) | Web-behavior-augmented recommendations | |
| KR100782047B1 (en) | Automatic Generation Method of the Initial Query Set for Image Search on the Mobile Web | |
| WO2015025248A2 (en) | A system apparatus circuit method and associated computer executable code for hybrid content recommendation | |
| Kim | Recommender system based on product taxonomy in e-commerce sites. |
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 |