CN103686237A - Method and system for recommending video resources - Google Patents

Abstract

The invention discloses a method and system for recommending video resources, including: collecting historical records in the process of watching videos by each user; making statistics on the collected historical records of each user, and calculating the number of videos of various types watched by each user. View the proportion of the number of videos of each type, and generate user groups including typical users corresponding to each specific type of video; respectively obtain the data characteristics of typical users in other dimensions in each specific type of user group, and determine the specific types of users in other dimensions. Group characteristics on the dimension; determine whether the data characteristics of other users other than typical users on other dimensions meet the group characteristics of a specific type of group, and if so, add the user to the specific type of user group; according to each user group The corresponding specific type information is used to recommend corresponding video resources to the users in the user group. Through the present invention, when a user watches a video, content recommendation can be carried out in a targeted manner, and the recommendation effect can be improved.

Description

Method and system for recommending video resources

technical field

The present invention relates to the field of smart TV technology, in particular to a method and system for recommending video resources.

Background technique

With the widespread intelligence of mobile phones and tablet computers, smart TVs are gradually entering our lives. Smart TVs, like smart phones, have a fully open platform and are equipped with an operating system. Users can install and uninstall software, games and other programs by themselves. Such programs may be provided by smart TV manufacturers or third-party service providers. Provided, through such programs, the functions of smart TVs can be continuously expanded. In addition, it is also possible to surf the Internet through a network cable or a wireless network. That is, a real smart TV can obtain program content from various channels such as the Internet, AV equipment, and PC, and through an easy-to-use integrated operation interface, the content most needed by consumers can be clearly displayed on the large screen through simple operations .

At present, there are generally only two ways for users to obtain resources from smart TVs. One is the most traditional way, that is, the user actively selects the information source through various access portals provided in the smart TV interface, and then clicks to watch. For example, a user may choose to watch a traditional TV signal source, or a program in a carousel channel provided by a smart TV server, and so on. The other is that the server pushes some recommended information to the user, such as pushing the latest popular videos to the user, and the user receives it passively.

By pushing some videos to the user, the user can obtain richer information through the smart TV. However, when the existing technology pushes video information, it mainly considers information such as the popularity of the video. After pushing to users, most users may not be interested in it, and then they are ignored by users, wasting the time spent in the push process. system and network resources.

Therefore, a technical problem that urgently needs to be solved by those skilled in the art is how to improve the effectiveness of recommendation during the process of pushing video information to users.

Contents of the invention

Aiming at the defects in the prior art, the present invention provides a method and system for recommending video resources.

A method for recommending video resources, comprising:

Collect historical records of each user watching videos; the historical records include the types of videos watched and data on at least one other dimension;

Make statistics on the collected historical records of each user, calculate the proportion of the number of videos watched by each user according to the number of videos of each type watched by each user, and classify the user into a specific type of video whose quantity ratio is greater than the group threshold The corresponding user group, the user is defined as a typical user, and a user group including typical users corresponding to each specific video type is generated;

Obtain the data characteristics of typical users in other dimensions in each specific type of user group respectively, and determine the group characteristics of each specific type of user in other dimensions;

Determine whether the data characteristics of users other than typical users in other dimensions meet the group characteristics of a specific type of user group, and if so, add the user to the specific type of user group;

According to the specific type information corresponding to each user group, corresponding video resources are recommended to users in the user group.

Optionally: the data on the at least one other dimension includes, the viewing duration of users watching various types of videos, the time period distribution of users watching videos, the number of times users watch the same video, the number of times users search and search content, user The number of times the application is used, and the number of times the user clicks on the recommended message.

Optionally: the collecting the history records of each user during watching the video includes: recording the history records of the users when the duration of watching the video is longer than the first lower limit.

Optionally: the collecting the historical records of each user during watching the video includes performing normalization processing on the collected historical records and deleting duplicate data.

Optionally: the group threshold is the second lower limit value of the proportion of the number of videos of various types watched by the user.

Optionally, it also includes: calculating the number of videos of each type watched by other users other than the typical user, calculating the number ratio of each type, and judging whether the number ratio is greater than or equal to the group threshold of the corresponding group, and if it is greater than or equal to the group threshold, the It is classified into this group; if the proportion is less than the group threshold of each group, the data characteristics of the user in other dimensions are obtained.

Optionally, the method further includes: statistically grouping the users whose history records are greater than the third lower limit, and re-grouping the grouped users after adding a certain number of records.

Optionally: the recommending corresponding videos to users in the user group includes recommending the same video across groups.

A system for recommending video resources, comprising:

A historical record acquisition module, configured to collect historical records of each user watching a video, the historical records including the types of videos watched, and at least one other dimension of data;

A typical user selects the division module, collects statistics on the historical records of each user, calculates the proportion of each type of video watched by each user according to the number of videos watched by each user, and classifies the user as a number whose proportion is greater than The user group corresponding to the specific type of video of the group threshold, the user is defined as a typical user, and a user group including typical users corresponding to each specific video type is generated;

The model training module is used to respectively obtain the data characteristics of typical users in other dimensions in each specific type of user group, and determine the group characteristics of the group in other dimensions;

The user grouping module is used to determine whether the data characteristics of other users other than typical users in other dimensions meet the group characteristics of a specific type of user group, and if so, add the user to the specific type of user group;

The video recommendation module is configured to recommend corresponding video resources to users in the user group according to the specific type information corresponding to each user group.

Optionally: the data on the at least one other dimension includes, the viewing duration of users watching various types of videos, the time period distribution of users watching videos, the number of times users watch the same video, the number of times users search and search content, user The number of times the application is used, and the number of times the user clicks on the recommended message.

Compared with the prior art, one of the above technical solutions has the following advantages or beneficial effects:

By counting the historical data of users watching videos, it is possible to determine which viewing category the user belongs to. By counting the data distribution of user data at different angles, it is possible to more accurately group users into categories. In this way, when users watch videos It is possible to carry out targeted content recommendation and improve the recommendation effect.

the

Description of drawings

Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiment. The drawings are only for the purpose of illustrating a preferred embodiment and are not to be considered as limiting the invention. Also throughout the drawings, the same reference numerals are used to designate the same components. In the attached picture:

Fig. 1 is the method flowchart of the recommended video resource provided by the embodiment of the present invention;

FIG. 2 is a schematic structural diagram of a device for recommending video resources provided by an embodiment of the present invention.

the

Detailed ways

Below in conjunction with accompanying drawing, further elaborate the present invention. It should be understood that these examples are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by the appended claims of the present application. the

The smart TV system of the present invention includes a client and a server, and users can access various video resources on the server after logging in to a specific account through the client. video resources on . The server saves the viewing history records of each account, and recommends video resources of interest to different users by analyzing the viewing history records. The present invention will be further introduced below through specific embodiments.

Example 1

As shown in FIG. 1 , the present invention firstly provides a method for recommending video resources. Video resources can be divided into action films, literary films, documentaries, animation films, etc. according to their types. During the video viewing process, due to the different viewing habits and personal preferences of each person, the proportion of the number of videos watched by each user is different. , if the user's historical viewing records can clearly reflect that the user's viewing ratio of a certain type of video is significantly higher than that of other types, then the user can be classified into the user group of this type. For example, users can be divided into action group, art group, documentary group, animation group and so on. Of course, the viewing records of some users may not be able to reflect the significance of a certain type directly from the proportion of the number of video types watched. At this time, how to determine the user group to which these users belong has become a key problem to be solved. question. The process of user division is described in detail below.

S101: Collect historical records of each user watching videos; the historical records include the types of videos watched and data in at least one other dimension;

In order to classify many users, it is first necessary to record the viewing history of the users. In specific implementation, after the user logs in to the server using his account, the server can receive the account information, judge the legitimacy of the account information, and when the account information is legal, collect the data of each video type that the user corresponding to the account has watched while watching the video. history record. Wherein, during specific implementation, the user may log in to the server by scanning a two-dimensional code, inputting a user ID and password, and other methods. The viewing history of the user may be collected by the client, and then uploaded to the server, so that the server can obtain the historical viewing records of each client. Among them, the client can upload in real time, or upload periodically (for example, upload once a week), or upload quantitatively (for example, upload once every recorded 10M data), and so on.

Among them, the historical records may include the number of various types of videos watched by the user, the viewing duration of each type of video, the time period distribution of the user watching the video; the number of times the user watches the same video; the number of times the user searches and the type distribution of the search content; The number of times the application is used and the type distribution of the application; the number and proportion of user clicks on recommended messages, etc.

The server can judge whether it is necessary to record the viewing behavior according to the length of time the user watches the video. Specifically, the server can record the user's viewing time in minutes. , the action is invalid and will not be recorded. This not only reduces the amount of data recording, but also records more accurately.

After the server receives the viewing history records of each user, it can firstly perform normalization processing on the data and delete duplicate data. Then, based on the processed historical records, the proportion of users watching various types of videos can be calculated. For example, there are 100 viewing records in the historical records. Among the 100 viewing records, the user watched 70 action movies, 10 literary movies, etc., then the calculated proportion of users watching action videos is 70%, literary and artistic videos accounted for 10% and so on.

S102: Make statistics on the collected historical records of each user, calculate the proportion of the number of videos watched by each user according to the number of videos of each type watched by each user, and classify the user into a specific group whose number ratio is greater than the group threshold The user group corresponding to the type of video, the user is defined as a typical user, and a user group including typical users corresponding to each specific video type is generated;

Among the users with existing historical records, some users have obvious classification characteristics. For example, some users watch more than 60% of action videos, some users watch more than 70% of literary videos, and some users watch animation videos. If the proportion of video is greater than 80%, etc., these users can be divided into action group, art group, animation group, etc. correspondingly. Other users do not have obvious characteristics. For example, some users have incomplete historical records, and there may be missing data. For example, some video types are difficult to classify, so the system cannot obtain the record of the video type in this viewing history. , or the network is interrupted during the viewing process and the viewing duration cannot be obtained, etc., or the classification of some users’ viewing types is not clear. It is difficult to recommend videos to these users.

In order to solve this problem, the present invention selects a part of users with obvious classification characteristics to divide into groups, and determines the typical users whose proportion of watching a specific type of video is greater than the group threshold, so as to prepare for the following model training. For example, if 1,000 users are selected, the proportion of the 1,000 users who watch the corresponding type of video is greater than the corresponding group threshold. For example, among the 1,000 users, the proportion of users who belong to the action group to watch action videos is greater than 60%, and those who belong to the art group The proportion of users watching literary and artistic videos is greater than 65%. Such users are called typical users, and here 60% and 65% are called group thresholds. The group thresholds of different groups can be the same or different, and the group thresholds can be adjusted dynamically.

S103: Obtain the data characteristics of typical users in other dimensions in each specific type of user group respectively, and determine the group characteristics of each specific type of user in other dimensions;

After completing the grouping of typical users, perform model training to obtain the data characteristics of each typical user in the corresponding group in other dimensions. The following uses the action category as an example to illustrate it. For example, among 1000 users, there are 100 users belonging to the action group, and the proportion of these 100 people watching action videos is greater than 60%, except that the proportion of watching video types is greater than 60%. The first data feature is that this group also has some data features in other dimensions. For example, the average viewing time of such users is more than 60 minutes, and the viewing time is often at night, and the average click rate on recommended messages is greater than 70%. In this example, the user's average movie watching time, movie watching time period, and the proportion of clicks on recommended messages are referred to as data on other dimensions.

Another example is that among the 1,000 users, there are 200 users belonging to the literature and art group. The proportion of these 200 people watching literature and art videos is greater than 65%. Data characteristics in other dimensions, for example, the viewing time of such users is often at noon, the number of times users watch the same video is more than 10 times, and the type of search content is more than 40% in the art category. In this example, users view Movie time period, the number of times users watch the same video, and the type distribution of search content are called data on other dimensions.

Through this model training, the data characteristics of each group of users in other dimensions can be obtained. Those skilled in the art should understand that in the collected historical viewing records, any data that is different from the proportion of watching each video type can be called It is data in other dimensions, and the embodiment of the present invention is not intended to limit the content included in the data in other dimensions.

S104: Determine whether the data characteristics of other users other than typical users in other dimensions meet the group characteristics of a specific type of user group, and if so, add the user to the specific type of user group;

After obtaining data characteristics in other dimensions, these data characteristics can be used to group users other than typical users. It is equivalent to using the trained model to classify users.

Specifically, it can be first judged whether the record about the type of video watched in the acquired historical records is complete, if the record is complete, then calculate the proportion of each video type it watches, and judge whether this proportion is greater than that of the corresponding group Group threshold, if it is greater than the group threshold, it will be classified into this group. When judging whether the records about the type of video they watched in the historical records are complete, you can set a lower limit value, and only for users whose missing data is greater than the lower limit value can further judge in other dimensions, for example, only for users whose missing records are greater than 10% For further judgments in other dimensions, this lower limit can be dynamically adjusted.

If the record is incomplete or its proportion is less than the group threshold of each group, the data characteristics of the user in other dimensions are obtained, and it is judged whether it meets the data characteristics of other dimensions in a certain type that are statistically obtained through model training. It is classified into this group. If it is found after the judgment that it does not exist in the group corresponding to the user, that is, the user cannot be classified into any type of group, then the user will be grouped into an ungrouped user.

For example, for a certain user, there are 500 records in the historical records, the system first judges whether the records about the types of videos watched by the user in the 500 historical records are complete, and if the records are complete, calculate the proportion of each video type watched by the user. If 400 of the 500 records are viewing action records, the proportion of the number is 80%, which is greater than the group threshold of 60% of the action group, then it will be classified into the action group; if there are 100 of the 500 records is a record of watching actions, or the record about the type of video watched in the 500 historical records is incomplete, then judge whether the data on the user’s said at least one other dimension satisfies the characteristics of a specific type , such as judging whether the user’s average viewing time is greater than 60 minutes, whether they often watch movies at night, and whether the proportion of clicks on recommended messages is greater than 70%. If these conditions are met, the user is classified into the action group; There is a group corresponding to the user, that is, the user cannot be classified into any type of group, then the user is grouped into an ungrouped user.

It should be understood that the same user in the present invention can be grouped into different groups. For example, for the same user, the data on other dimensions may simultaneously satisfy the data characteristics on the other dimensions in multiple corresponding grouping models, and the user is divided into groups at the same time. into these two groups.

In the present invention, as the viewing time of users increases, the information of their historical records becomes more and more abundant. In order to realize the grouping of users more accurately, the system only performs statistical grouping for users whose historical records are greater than the lower limit. , the lower limit may be 10, for example. After grouping, the system can re-group the users after adding a certain number of records. For example, after adding 10 records, the system will re-group the original records and the 10 new records.

S105: According to the specific type information corresponding to each user group, recommend corresponding video resources to users in the user group.

After the user grouping is completed, after the user logs in to the system through the account, the history record can be associated with the user account, and the user's viewing habits can be analyzed, such as watching TV series, watching movies, watching domestic or American dramas for TV series, and watching sci-fi movies for movies. , horror movies or romance movies, etc., recommend specific types of video resources to users according to the group information corresponding to each user. For example, if an account recorded by the server belongs to the action group, the system will automatically recommend action videos to the user after logging in through the account. The same video can be recommended across groups. For example, for a video with both action and comedy features, the video is recommended to the action combination comedy group at the same time, preferably to the intersection users of the two groups.

Videos can be recommended to users in the form of posters or messages. The poster screen will be displayed when the client is turned on. For example, if the client is a TV, related videos recommended to users will be displayed on the TV startup screen. The screen contains links, and the user clicks on the screen You can directly link to related videos to watch; or recommend to users by popping up messages during the process of watching videos. The messages include video names, video addresses, thumbnails, etc.

When the user does not log in through the account or the user is a newly registered user or the user belongs to an ungrouped user group, the system can recommend other videos of this type to him according to the type of video he is currently watching, so that it is convenient for the user to watch similar videos, or random recommendation.

Example 2

As shown in FIG. 2 , the present invention also provides a system for recommending video resources. The system includes a client and a server. Users can access various video resources on the server after logging in to a specific account through the client. The client is an application program running on a smart TV terminal. Users can watch video resources on the server through the terminal device. The server saves the viewing history records of each account, and recommends video resources of interest to different users by analyzing the viewing history records.

A historical record acquisition module 201, configured to collect historical records of each user watching a video, the historical records including the types of videos watched, and data on at least one other dimension;

A typical user selects the division module 202, collects statistics on the historical records of each user, calculates the proportion of the number of videos watched by each user according to the number of various types of videos watched by each user, and classifies the user into the number proportion A user group corresponding to a specific type of video greater than the group threshold, the user is defined as a typical user, and a user group including typical users corresponding to each specific video type is generated;

The model training module 203 is used to obtain the data characteristics of typical users in other dimensions in each specific type of user group, and determine the group characteristics of the group in other dimensions;

The user grouping module 204 is used to judge whether the data characteristics of other users outside the typical user in other dimensions meet the group characteristics of a certain type of user group, and if so, add the user to the specific type of user group;

The video recommendation module 205 is configured to recommend corresponding video resources to users in the user group according to the specific type information corresponding to each user group.

Wherein, the data on the at least one other dimension includes the viewing time of various types of videos that users watch, the time period distribution of users watching videos, the number of times users watch the same video, the number of times users search and what they search for, and the number of times users use applications. The number of times, and the number of user clicks on recommended messages.

During specific implementation, the historical record acquisition module is specifically configured to: record the user's historical record when the duration of watching the video by the user is greater than the first lower limit value. The group threshold is the second lower limit value of the proportion of the number of videos of various types watched by the user.

In order to reduce the amount of computation, the system can also include:

The normalization processing module is used for normalizing the collected historical records and deleting duplicate data.

The first judging module is used to calculate the quantity of each type of video watched by other users other than the typical user, calculate the quantity proportion of each type, and judge whether the quantity proportion is greater than or equal to the group threshold of the corresponding group. If it is greater than or equal to the group threshold, then the It is classified into the group; if the proportion is less than the group threshold of each group, the user grouping module is triggered to perform the operation of judging the data characteristics in other dimensions.

The regrouping module is used to perform statistical grouping on users whose history record number is greater than the third lower limit, and perform statistical grouping on the grouped users every time a certain number of records is added.

For the aforementioned device embodiments, for the sake of simple description, they are expressed as a series of module combinations, but those skilled in the art should know that the present invention is not limited by the described module combinations, because according to the present invention , some modules can be implemented by other modules; secondly, those skilled in the art should also know that the above device embodiments are all preferred embodiments, and the modules involved are not necessarily required by the present invention.

Each embodiment in this specification is described in a progressive manner, each embodiment focuses on the difference from other embodiments, and the same and similar parts of each embodiment can be referred to each other. As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and for the related parts, please refer to the part of the description of the method embodiment.

The method and system for recommending video resources provided by the present invention are described above in detail. In this paper, specific examples are used to illustrate the principle and implementation of the present invention. The descriptions of the above embodiments are only used to help understand the present invention. method and its core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation and application scope. Invention Limitations.

Claims (10)

1. a method of recommending video resource, is characterized in that, comprising:

Collect each user and watch the historical record in video process; Described historical record comprises each video type of watching, and the data at least one other dimension;

Historical record to each user who collects is added up, the quantity of all types of videos of watching according to each user is calculated its quantity accounting of watching all types of videos, and this user is included into the corresponding user's group of particular type video that quantity accounting is greater than group threshold value, this user is defined as typical user, generates the user's group that comprises typical user corresponding to each particular video frequency type;

The data characteristics of the user Zu Nei typical user that obtains respectively each particular type in other dimensions, the stack features of the user who determines each particular type in other dimensions;

Whether the data characteristics of judgement other users outside typical user in other dimensions, meet the stack features of user's group of certain particular type, if so, this user added to user's group of this particular type;

According to each user, organize corresponding particular type of information, the user in user's group recommends corresponding video resource.

2. method according to claim 1, it is characterized in that: the data in described at least one other dimension comprise, user watches watch duration, the user of all types of videos to watch number of times that the number of times that the time period distributes, user watches same video of video, number of times that user searches for and search content, user use application and user to recommending the number of clicks of the message of class.

3. method according to claim 1, is characterized in that: each user of described collection watches the historical record in video process to comprise: when user watches the duration of video to be greater than the first lower limit, record its historical record.

4. method according to claim 1, is characterized in that: each user of described collection watches the historical record in video process to comprise, the historical record of collecting is normalized to deleting duplicated data.

5. method according to claim 1, is characterized in that: the second lower limit of the quantity accounting of all types of videos that described group of threshold value watched for user.

6. method according to claim 1, it is characterized in that, also comprise: the quantity of all types of videos that other users outside calculating typical user watch is calculated all types of quantity accountings, judge whether this quantity accounting is more than or equal to the group threshold value of corresponding group, if be more than or equal to group threshold value, put under this group; If accounting is less than the group threshold value of each grouping, obtain the data characteristics of this user in other dimensions.

7. method according to claim 1, it is characterized in that, also comprise: the user who is greater than the 3rd lower limit for historical record number carries out statistical packet, for the user after grouping, after the record of the certain number of every increase, again this user is carried out to statistical packet.

8. method according to claim 1, is characterized in that: described user in user's group recommends corresponding video to comprise same video is carried out recommending across group.

9. a system of recommending video resource, is characterized in that, comprising:

Historical record acquisition module, watches the historical record of video process for collecting each user, described historical record comprises each video type of watching, and the data at least one other dimension;

Typical user chooses division module, historical record to each user who collects is added up, the quantity of all types of videos of watching according to each user is calculated its quantity accounting of watching all types of videos, and this user is included into the corresponding user's group of particular type video that quantity accounting is greater than group threshold value, this user is defined as typical user, generates the user's group that comprises typical user corresponding to each particular video frequency type;

Model training module, the data characteristics for the user Zu Nei typical user that obtains respectively each particular type in other dimensions, determines the stack features of this group in other dimensions;

Whether user grouping module, for judging other users outside typical user data characteristics in other dimensions, meet the stack features of user's group of certain particular type, if so, this user added to user's group of this particular type;

Video recommending module, for organizing corresponding particular type of information according to each user, the user in user's group recommends corresponding video resource.

10. system according to claim 9, it is characterized in that: the data in described at least one other dimension comprise, user watches watch duration, the user of all types of videos to watch number of times that the number of times that the time period distributes, user watches same video of video, number of times that user searches for and search content, user use application and user to recommending the number of clicks of the message of class.

Images