WO2017177643A1

WO2017177643A1 - Multimedia recommendation method and device

Info

Publication number: WO2017177643A1
Application number: PCT/CN2016/102735
Authority: WO
Inventors: 孙浩川
Original assignee: 乐视控股（北京）有限公司; 乐视网信息技术（北京）股份有限公司
Priority date: 2016-04-12
Filing date: 2016-10-20
Publication date: 2017-10-19
Also published as: CN105843953A

Abstract

A multimedia recommendation method and device. The method comprises: extracting feature data of information of data to be processed in a database (S110); carrying out feature construction on the feature data to obtain target feature data (S120); calculating a recommendation score corresponding to each multimedia in the database according to the target feature data (S130); and recommending the multimedia to a user according to the recommendation score (S140). In this way, the recommendation score of each multimedia for a user in a database is finally calculated according to comprehensive feature data of the user such as interest characteristics and historical behaviors, and the multimedia with a high recommendation score is recommended to the user, thereby recommending to each user multimedia information of interest to the user in a targeted manner.

Description

Multimedia recommendation method and device

The present application claims the priority of the Chinese Patent Application, the entire disclosure of which is hereby incorporated by reference.

Technical field

The embodiments of the present invention relate to the field of video technologies, and in particular, to a multimedia recommendation method and apparatus.

Background technique

With the continuous development of technology, there are more and more types and types of mobile terminal products that can be provided to users, such as smart phones, tablets and notebook computers. These mobile terminal products almost all have wireless network WIFI access. Function, users can easily connect the terminal to WIFI to watch videos online. In order to facilitate users to quickly find their favorite videos, many video applications provide users with video recommendation functions.

At present, based on the video viewing behavior of all users, it is recommended to the user to find other videos that have been viewed by the same type of users for each video. Although this method can mine the video that the user may be interested in according to the user's viewing video history, but only based on the video, the recommendation is not comprehensively considered, and the user's historical data is simply counted. The calculation of the probability that the user can click on the video is optimized, and the recommended video may not be the user's favorite video.

Summary of the invention

In order to solve the related art, when recommending the video that the user likes, the recommendation is based on the video only, and the user's interest is not comprehensively considered. The recommended video is probably not the problem of the video that the user likes. The present invention provides a multimedia. Recommended methods and devices.

According to a first aspect of the embodiments of the present invention, a multimedia recommendation method is provided, including:

Extracting feature data of data to be processed in the database;

Performing feature construction on the feature data to obtain target feature data;

Calculating a recommendation score corresponding to each multimedia in the database according to the target feature data;

The multimedia is recommended to the user based on the recommendation score.

According to a second aspect of the embodiments of the present invention, a multimedia recommendation apparatus is provided, including:

a feature data extracting unit, configured to extract feature data of the data to be processed in the database;

a feature construction unit configured to perform feature construction on the feature data to obtain target feature data;

a recommendation score calculation unit, configured to calculate, according to the target feature data, a recommendation score corresponding to each multimedia in the database;

And a video recommendation unit, configured to recommend the multimedia to the user according to the recommended score.

The embodiment of the present application provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium can store a computer program, and the program can implement a part of the multimedia recommendation method described above or All steps.

An embodiment of the present application further provides an electronic device, including: one or more processors; and a memory; wherein the memory stores instructions executable by the one or more processors, the instructions being The one or more processors are executed to enable the one or more processors to perform the multimedia recommendation method described above.

An embodiment of the present application provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions, when the program instructions are executed by a computer, The computer is caused to execute the above multimedia recommendation method in the embodiment of the present application.

The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

The multimedia recommendation method and device provided in the embodiment of the present invention extract feature data of the data to be processed in the database, and perform feature construction on the feature data to obtain target feature data, and calculate each multimedia in the database according to the target feature data. The corresponding recommendation score, based on the recommendation score, recommends multimedia for the user. Since the feature data extracted from the to-be-processed data information may include multimedia related information and user's own behavior information, the extracted feature data is more comprehensive, and finally, each of the databases may be finally calculated according to more comprehensive feature data. The recommended scores of multimedia are more in line with the user's actual preference. Because each user's feature data is different, the multimedia calculated for each user The body recommendation scores are also different, so that it is possible to specifically recommend multimedia information of its own interest to each user. After recommending the multimedia with higher recommendation score to the user, the user experience can be more satisfied and the user experience can be improved.

The above general description and the following detailed description are intended to be illustrative and not restrictive.

DRAWINGS

The accompanying drawings, which are incorporated in the specification of FIG

FIG. 1 is a flowchart of a multimedia recommendation method according to an embodiment of the present invention;

FIG. 2 is a specific flowchart of characterizing the feature data to obtain target feature data in the embodiment;

FIG. 3 is a specific flowchart of calculating a recommendation score corresponding to each multimedia in the database according to the target feature data in the embodiment;

FIG. 4 is a specific flowchart of recommending multimedia for a user according to a recommendation score in the embodiment;

FIG. 5 is a schematic structural diagram of a multimedia recommendation apparatus according to another embodiment of the present invention;

Figure 6 shows a schematic view of a feature construction unit in another embodiment;

Figure 7 is a diagram showing a recommended score calculation unit in another embodiment;

Figure 8 is a diagram showing a video recommendation unit in another embodiment;

FIG. 9 is a schematic diagram showing the hardware structure of the device of the multimedia recommendation method provided by the embodiment of the present application.

detailed description

Exemplary embodiments will be described in detail herein, examples of which are illustrated in the accompanying drawings. The following description refers to the same or similar elements in the different figures unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with aspects of the invention as detailed in the appended claims.

The embodiment of the present invention first provides a multimedia recommendation method, as shown in FIG. 1 , the method may include Including the following steps:

In step S110, feature data of the data information to be processed in the database is extracted.

It should be noted that the multimedia involved in the embodiment of the present invention may be data such as video, audio, or image. For convenience of description, the multimedia involved in the embodiment of the present invention is illustrated by using video as an example, and may also be applied. The audio, image, and the like that the user likes can be recommended, for example, in the embodiment of the present invention.

The data information to be processed here includes information about the video stored in the database, information about the user, and behavior records of the user watching the video.

Extracting feature data from the to-be-processed data information in the above, for example, extracting feature data of the video from the database, including: the channel to which the video belongs, the type of the video, the region to which the video belongs, the release time of the video, the update time, and the video The click rate of a certain period of time in the past; the feature data of the user's information extracted from the database includes: the user's age, gender, occupation, the channel that has watched the video, the type of video that has been watched, the TAG (tag) that has watched the video, And the area where the video has been viewed, etc.; the behavior record of the user watching the video may be a historical behavior of the user viewing the historical video from the database for a certain period of time, and the characteristics include: the user's identification code, a list of recommended videos, a list, Recommended time, and whether you have clicked on a recommended video.

In addition, in the process of selecting feature data, it is necessary to remove data that is lost in some fields, that is, invalid data. The video information in the video library can be divided into two categories, one is a video that the user likes to watch, and the other is a video that the user does not like to watch. The video that the user likes to watch can be used as a positive sample, and the video that the user does not like to watch can be regarded as a negative sample. By dividing the video information in the video library into positive and negative samples, the accuracy of the subsequent model fitting can be improved. So that after getting a video, you can more accurately identify whether the video is a video that the user likes.

Exemplarily, in the video recommended by the user in the past, the video that the user clicked is taken as a positive sample, and the video that the user has not clicked is taken as a negative sample. Therefore, in order to ensure the uniformity of the positive and negative samples, it is necessary to randomly sample a certain amount of data in the negative sample and extract its characteristics, so that the ratio of the positive and negative samples is 1:1.

In step S120, the feature data is subjected to feature construction to obtain target feature data.

The feature data directly extracted from the data to be processed can not directly meet the requirements of model training. Therefore, it is necessary to first construct the feature data, for example, to discretize feature data from continuous features; Encoding to convert feature data from one-dimensional features to multidimensional Features, each feature corresponds to a feature value; in order to express complex nonlinear relationships, multi-dimensional feature data is intersected by feature intersection, feature data after feature construction, and then model training can obtain more accurate results.

It should be noted that, in the process of performing model training on the feature data after feature construction, each pair of models is trained once, which is equivalent to extracting feature data from the data to be processed once, and identifying the feature data once, and the feature data is Classified into a certain category, such as the video corresponding to the feature data is classified as a video that the user likes or a video that the user does not like. The more times you train, the more accurate the recognition. When identifying whether a video is a video that the user likes, the extracted feature of the video can be quickly and accurately determined to identify that the video is a video that the user likes or that is not.

In step S130, a recommendation score corresponding to each multimedia in the database is calculated according to the target feature data.

In the process of calculating the recommendation score corresponding to each video in the database by using the target feature data, the FTRL algorithm may be first used to fit the obtained feature data, and finally a logistic regression model that can express the user's preference for the video is obtained. The logistic regression model is represented by a KV pair, that is, a key-value, where the key identifies the feature name (such as a video name) and the value is a feature weight (such as a recommended score).

In step S140, the user is recommended for the multimedia based on the recommendation score.

In the embodiment, the key obtained above may be corresponding to a specific video name, and the value corresponds to a recommended score corresponding to the video name. By training the feature data, a recommendation score for each video in the database can be obtained, the recommendation score can be sorted, and a video with a higher recommendation score is recommended to the user.

The multimedia recommendation method provided in the embodiment of the present invention extracts feature data of the data to be processed in the database, and performs feature construction on the feature data to obtain target feature data, and calculates each multimedia corresponding to the database according to the target feature data. A recommendation score based on which the user is recommended for multimedia. Since the feature data extracted from the to-be-processed data information may include multimedia related information and user's own behavior information, the extracted feature data is more comprehensive, and finally, each of the databases may be finally calculated according to more comprehensive feature data. The recommended scores of multimedia are more in line with the user's actual preference. Since the feature data of each user is different, the multimedia recommendation score calculated for each user is also different, thereby achieving targeted recommendation of multimedia information of its own interest to each user. After recommending the multimedia with higher recommendation score to the user, the user experience can be more satisfied and the user experience can be improved.

In order to elaborate the feature structure of the extracted feature data, as a refinement of the method of FIG. 1, in another embodiment of the present invention, as shown in FIG. 2, step S120 may further include the following steps:

In step S121, the feature data is discretized to obtain discretized feature data.

When using the logistic regression model to train the feature data, since the logistic regression model is a relatively simple linear model, it can not identify the continuous values in the feature well, so it is necessary to discretize the continuous features, such as a certain segment of the video. The time-click rate is generally divided by equal frequency. It is to sort all the samples on the feature that needs to be discretized, and divide it into several equal parts, and replace the value of the original feature with its index.

In step S122, the discretized feature data is encoded to generate multi-dimensional discretized feature data.

Since the logistic regression model is a relatively simple linear model, the linear model can only represent two-dimensional class features, but not multi-dimensional class features. Illustratively, the types of videos can be divided into various types according to the plot, such as: war, love, and life. Therefore, it is necessary to encode the obtained discretized feature data, and convert the original one-dimensional feature data into multi-dimensional feature data, so that the obtained multi-dimensional feature data can be more comprehensive, and the final recommended score is more suitable for the user's preference. For example, the n-dimensional feature data is transformed, where n is the corresponding value in the category feature data. The one-hot encoding method may be adopted when encoding the discretized feature data, and the embodiment is not limited thereto.

In step S123, the multi-dimensional discretized feature data is subjected to feature intersection according to a predetermined rule, and converted into target feature data.

Since the logistic regression model is a relatively simple linear model and cannot express complex nonlinear relationships, it is necessary to manually formulate rules for feature cross. In an exemplary embodiment, the user is mainly based on the characteristics of LeTV video. Intersect with the characteristics of the video. Since the feature intersection requires a large amount of computation time, here considering the performance problem, the feature cross is three-dimensional at most, as shown in Table 1.

Table 1

序号Serial number	第一维First dimension	第二维Second dimension	第三维Third dimension
11	用户感兴趣频道User interest channel	视频所属频道Video belongs to the channel
22	用户感兴趣类型User interest type	视频类型Video type
33	用户感兴趣类型User interest type	视频TAGVideo TAG
44	用户感兴趣TAGInterested in TAG	视频TAGVideo TAG
55	用户感兴趣地区User area of interest	视频地区Video area
66	用户感兴趣类型User interest type	视频类型Video type	视频发布时间Video release time

77	用户感兴趣类型User interest type	视频类型Video type	推荐时间Recommended time
88	用户感兴趣类型User interest type	视频频道Video channel	推荐时间Recommended time

As a refinement of the method of FIG. 1, in another embodiment of the present invention, as shown in FIG. 3, step S130 may further include the following steps:

In step S131, feature encoding is performed on the target feature data to obtain integer feature data.

Since the above generated features are all represented by a string, the form of the string can be used to facilitate the user to view the meaning of the representation, but for the computer, the operation speed of the string is greatly reduced, especially in the case of massive data processing. , will seriously reduce the computer's computing speed. Therefore, in order to facilitate efficient calculation of the computer, in order to obtain the operation result quickly, in the embodiment of the present invention, the order of occurrence of each feature data obtained is encoded from 0, and the feature data of the string type is converted into integer data, such as Int type data, which is convenient for quick calculation results.

In step S132, the integer feature data is fitted to obtain a corresponding data model.

In step S133, the weight corresponding to each of the integer feature data is read in the data model, and the weights are accumulated to obtain a recommendation score corresponding to each multimedia.

For the process of fitting the integer feature data, the FTRL algorithm can be used to fit the integer feature data, so that the corresponding data model is obtained by fitting. The data module is a logistic regression model. The representation of the logistic regression model is a K-V pair, that is, K is a key, indicating a feature name, such as a video name, and V is a value, indicating a weight of the corresponding feature, which is a specific value. In the embodiment, the weight may be a recommendation score, and the higher the recommendation score, the higher the degree of preference for the user; on the contrary, the lower the degree the user likes. Therefore, the video with higher recommendation score can be recommended to the user according to the recommendation score in the logistic regression model, so that the video recommended for the user is likely to be the video that the user likes.

As a refinement of the method of FIG. 1, in another embodiment of the present invention, as shown in FIG. 4, step S140 may further include the following steps:

In step S141, the recommended scores are sorted in descending order to obtain a video recommendation list.

In step S142, a video in which the recommended score is greater than the preset value in the video recommendation list is used as the recommended video.

After obtaining the recommended score of each video in the database, according to the recommended score, the videos are sorted in descending order to obtain a video recommendation list. A video with a recommended score greater than a preset value in the video recommendation list is used as the recommended video. Of course, you can also arrange the first few videos in the video recommendation list. The higher video is recommended to the user. The embodiment of the invention is not limited thereto.

Through the description of the above method embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus a necessary general hardware platform, and of course, can also be through hardware, but in many cases, the former is better. Implementation. Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, may be embodied in the form of a software product stored in a storage medium, including a plurality of instructions for causing a A computer device (which may be a personal computer, server, or network device, etc.) performs all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes various types of media that can store program codes, such as a read only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

In addition, as an implementation of the foregoing embodiments, the embodiment of the present invention further provides a multimedia recommendation device, where the device is located in a terminal, as shown in FIG. 5, the device includes:

The feature data extracting unit 10 is configured to extract feature data of the data to be processed in the database;

The feature construction unit 20 is configured to perform feature construction on the feature data to obtain target feature data.

a recommendation score calculation unit 30, configured to calculate, according to the target feature data, a recommendation score corresponding to each multimedia in the database;

The video recommendation unit 40 is configured to recommend multimedia to the user according to the recommended score.

In another embodiment of the present invention, based on FIG. 5, as shown in FIG. 6, the feature construction unit 20 includes:

The feature discretization processing module 21 is configured to discretize the feature data to obtain a discretization feature data;

The encoding module 22 is configured to encode the discretized feature data to generate multi-dimensional discretized feature data;

The target feature data generating module 23 is configured to cross-divide the multi-dimensional discretized feature data into features according to a predetermined rule and convert the image into the target feature data.

In another embodiment of the present invention, based on FIG. 5, as shown in FIG. 7, the recommended score calculation unit 30 includes:

The feature encoding module 31 is configured to perform feature encoding on the target feature data to obtain integer feature data.

a data fitting module 32, configured to fit the integer feature data to obtain a corresponding data model;

The recommended score determining module 33 is configured to read the weight corresponding to each of the integer data features in the data model, and accumulate the weights to obtain a recommendation score corresponding to each multimedia.

In another embodiment of the present invention, based on FIG. 5, as shown in FIG. 8, the video recommendation unit 40 includes:

a sorting module 41, configured to sort the recommended scores in descending order to obtain a video recommendation list;

The recommended video determining module 42 is configured to use a video in the video recommendation list that has a recommended score greater than a preset value as the recommended video.

The feature data in the embodiment includes: video feature data, user feature data, and historical record feature data of the user viewing video.

With regard to the apparatus in the above embodiments, the specific manner in which the respective modules perform the operations has been described in detail in the embodiment relating to the method, and will not be explained in detail herein.

The multimedia recommendation device provided in the embodiment of the present invention extracts feature data of the data to be processed in the database, and performs feature construction on the feature data to obtain target feature data, and calculates each multimedia corresponding to the database according to the target feature data. A recommendation score based on which the user is recommended for multimedia. Since the feature data extracted from the to-be-processed data information may include multimedia related information and user's own behavior information, the extracted feature data is more comprehensive, and finally, each of the databases may be finally calculated according to more comprehensive feature data. The recommended scores of multimedia are more in line with the user's actual preference. Because each user's feature data is different, the multimedia recommendation calculated for each user The scores are also different, so that it is possible to specifically recommend multimedia information of its own interest to each user. After recommending the multimedia with higher recommendation score to the user, the user experience can be more satisfied and the user experience can be improved.

The embodiment of the present application further provides a non-transitory computer readable storage medium, where the non-transitory computer readable storage medium can store a program, and when executed, the program can implement a multimedia recommendation provided by any one of the foregoing embodiments. Some or all of the steps in the various implementations of the method.

FIG. 9 is a schematic structural diagram of hardware of an electronic device according to a multimedia recommendation method according to an embodiment of the present disclosure. As shown in FIG. 9, the device includes:

One or more processors 910 and memory 920, one processor 910 is taken as an example in FIG.

The apparatus for performing the multimedia recommendation method may further include: an input device 930 and an output device 940.

The processor 910, the memory 920, the input device 930, and the output device 940 may be connected by a bus or other means, as exemplified by a bus connection in FIG.

The memory 920 is used as a non-transitory computer readable storage medium, and can be used to store a non-volatile software program, a non-volatile computer executable program, and a module, such as a program instruction corresponding to the multimedia recommendation method in the embodiment of the present application. Module. The processor 910 executes various functional applications and data processing of the electronic device by executing non-volatile software programs, instructions, and modules stored in the memory 920, that is, implementing the multimedia recommendation method of the above method embodiments.

The memory 920 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created according to usage of the multimedia recommendation device, and the like. Moreover, memory 920 can include high speed random access memory, and can also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some embodiments, memory 920 can optionally include memory remotely located relative to processor 910, which can be connected to the multimedia recommendation device over a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

Input device 930 can receive input numeric or character information and generate key signal inputs related to user settings and function control of the multimedia recommendation device. Output device 940 can include a display device such as a display screen.

The one or more modules are stored in the memory 920 when the one or more processors are When the 910 is executed, the multimedia recommendation method in any of the above method embodiments is executed.

The above products can perform the methods provided by the embodiments of the present application, and have the corresponding functional modules and beneficial effects of the execution method. For technical details that are not described in detail in this embodiment, reference may be made to the method provided by the embodiments of the present application.

The electronic device of the embodiment of the invention exists in various forms, including but not limited to:

(1) Mobile communication devices: These devices are characterized by mobile communication functions and are mainly aimed at providing voice and data communication. Such terminals include: smart phones (such as iPhone), multimedia phones, functional phones, and low-end phones.

(2) Ultra-mobile personal computer equipment: This type of equipment belongs to the category of personal computers, has computing and processing functions, and generally has mobile Internet access. Such terminals include: PDAs, MIDs, and UMPC devices, such as the iPad.

(3) Portable entertainment devices: These devices can display and play multimedia content. Such devices include: audio, preview players (such as iPod), handheld game consoles, e-books, and smart toys and portable car navigation devices.

(4) Server: A device that provides computing services. The server consists of a processor, a hard disk, a memory, a system bus, etc. The server is similar to a general-purpose computer architecture, but because of the need to provide highly reliable services, processing power and stability High reliability in terms of reliability, security, scalability, and manageability.

(5) Other electronic devices with data interaction functions.

It will be appreciated that the present invention is applicable to a wide variety of general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor based systems, set-top boxes, programmable consumer electronics devices, network PCs, small computers, mainframe computers, including A distributed computing environment of any of the above systems or devices, and the like.

The invention may be described in the general context of computer-executable instructions executed by a computer, such as a program module. Generally, program modules include routines, programs, objects, components, data structures, and the like that perform particular tasks or implement particular abstract data types. The invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media including storage devices.

It should be noted that, in this context, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply these There is any such actual relationship or order between entities or operations. Furthermore, the term "comprises" or "comprises" or "comprises" or any other variations thereof is intended to encompass a non-exclusive inclusion, such that a process, method, article, or device that comprises a plurality of elements includes not only those elements but also Other elements, or elements that are inherent to such a process, method, item, or device. An element that is defined by the phrase "comprising a ..." does not exclude the presence of additional equivalent elements in the process, method, item, or device that comprises the element.

Other embodiments of the invention will be apparent to those skilled in the <RTIgt; The present application is intended to cover any variations, uses, or adaptations of the present invention, which are in accordance with the general principles of the present invention and include common general knowledge or conventional technical means in the art that are not disclosed in the present invention. . The specification and examples are to be considered as illustrative only,

It is to be understood that the invention is not limited to the details of the details of The scope of the invention is limited only by the appended claims.

Claims

A multimedia recommendation method, comprising:

Extracting feature data of data to be processed in the database;

Performing feature construction on the feature data to obtain target feature data;

Calculating a recommendation score corresponding to each multimedia in the database according to the target feature data;

The multimedia is recommended to the user based on the recommendation score.
The multimedia recommendation method according to claim 1, wherein the characterizing the feature data comprises:

Discretizing the feature data to obtain discretized feature data;

Encoding the discretized feature data to generate multi-dimensional discretized feature data;

The multi-dimensional discretized feature data is subjected to feature intersection according to a predetermined rule, and converted into the target feature data.
The multimedia recommendation method according to claim 1, wherein the calculating a recommendation score corresponding to each multimedia in the database according to the target feature data comprises:

Feature encoding the target feature data to obtain integer feature data;

Fitting the integer feature data to obtain a corresponding data model;

The weight corresponding to each of the integer feature data is read in the data model, and the weights are accumulated to obtain a recommendation score corresponding to each multimedia.
The multimedia recommendation method according to claim 1, wherein the recommending a video for the user according to the recommended score comprises:

Sorting the recommended scores in descending order to obtain a video recommendation list;

A video in which the recommended score of the video recommendation list is greater than a preset value is used as the recommended video.
The multimedia recommendation method according to any one of claims 1 to 4, wherein the feature data comprises: video feature data, user feature data, and history record feature data of the user viewing video.
A multimedia recommendation device, comprising:

a feature data extracting unit, configured to extract feature data of the data to be processed in the database;

a feature construction unit configured to perform feature construction on the feature data to obtain target feature data;

a recommendation score calculation unit, configured to calculate, according to the target feature data, a recommendation score corresponding to each multimedia in the database;

And a video recommendation unit, configured to recommend the multimedia to the user according to the recommended score.
The multimedia recommendation device according to claim 6, wherein the feature construction unit comprises:

a feature discretization processing module, configured to discretize the feature data to obtain discretized feature data;

An encoding module, configured to encode the discretized feature data to generate multi-dimensional discretized feature data;

And a target feature data generating module, configured to cross-divide the multi-dimensional discretized feature data into a target rule according to a predetermined rule.
The multimedia recommendation device according to claim 6, wherein the recommendation score calculation unit comprises:

a feature coding module, configured to perform feature coding on the target feature data to obtain integer feature data;

a data fitting module, configured to fit the integer feature data to obtain a corresponding data model;

And a recommendation score determining module, configured to read weights corresponding to each of the integer data features in the data model, and accumulate the weights to obtain a recommendation score corresponding to each multimedia.
The multimedia recommendation device according to claim 6, wherein the video recommendation unit comprises:

a sorting module, configured to sort the recommended scores in descending order to obtain a video recommendation list;

The recommended video determining module is configured to use a video in the video recommendation list that has a recommended score greater than a preset value as the recommended video.
The multimedia recommendation device according to any one of claims 6 to 9, wherein the feature data comprises: video feature data, user feature data, and history record feature data of the user viewing video.
A non-transitory computer readable storage medium, characterized in that the non-transitory computer readable storage medium stores a computer program for causing the computer to perform the method of any of claims 1-5 .
An electronic device comprising:

One or more processors; and,

Memory; characterized in that

The memory stores instructions executable by the one or more processors, the instructions being executed by the one or more processors to cause the one or more processors to:

Extracting feature data of data to be processed in the database;

Performing feature construction on the feature data to obtain target feature data;

Calculating a recommendation score corresponding to each multimedia in the database according to the target feature data;

The multimedia is recommended to the user based on the recommendation score.
A computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions that, when executed by a computer, cause the computer to execute The method of claims 1-5.