CN116070014A - Recommendation method, recommendation device, server and computer readable storage medium - Google Patents
Recommendation method, recommendation device, server and computer readable storage medium Download PDFInfo
- Publication number
- CN116070014A CN116070014A CN202111302296.9A CN202111302296A CN116070014A CN 116070014 A CN116070014 A CN 116070014A CN 202111302296 A CN202111302296 A CN 202111302296A CN 116070014 A CN116070014 A CN 116070014A
- Authority
- CN
- China
- Prior art keywords
- user
- content
- offline
- data
- tag
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/21—Design, administration or maintenance of databases
- G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/906—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Abstract
The embodiment of the application relates to the technical field of communication, and discloses a recommendation method, a recommendation device, a recommendation server and a computer readable storage medium. The recommendation method comprises the following steps: under the condition that the user is determined to be online, acquiring a user tag built offline for the user from an offline database; performing tag matching according to the user tag and the content tag which is stored in the offline database and is constructed offline for each content to be recommended, and generating a recommendation list for the user online according to a matching result; and pushing the recommendation list to the terminal corresponding to the user, so that the calculation efficiency can be improved, and the recommendation response speed can be accelerated.
Description
Technical Field
The embodiment of the application relates to the technical field of communication, in particular to a recommendation method, a recommendation device, a recommendation server and a computer readable storage medium.
Background
Along with the rapid development of the internet, information explosion becomes a normal state, and each large video recommendation platform can conduct targeted personalized recommendation on each user for increasing user viscosity, so that the requirement on the capability of a server side for processing data is higher and higher. The method aims at solving the problems that the calculation efficiency is low and the recommendation response speed is low when the current personalized recommendation system faces mass data.
Disclosure of Invention
The main purpose of the embodiments of the present application is to provide a recommendation method, a recommendation device, a server and a computer readable storage medium, so that the calculation efficiency can be improved, and the recommendation response speed can be accelerated.
To achieve at least the above object, an embodiment of the present application provides a recommendation method, including: under the condition that the user is determined to be online, acquiring a user tag built offline for the user from an offline database; performing tag matching according to the user tag and the content tag which is stored in the offline database and is constructed offline for each content to be recommended, and generating a recommendation list for the user online according to a matching result; and pushing the recommendation list to the terminal corresponding to the user.
To achieve at least the above object, an embodiment of the present application further provides a recommendation device, including: the off-line user tag acquisition module is used for acquiring the user tag constructed offline for the user from an off-line database under the condition that the user is determined to be on-line; the online matching module is used for carrying out tag matching according to the user tag and the content tag which is stored in the offline database and is constructed for each content to be recommended offline, and generating a recommendation list for the user online according to a matching result; and the online pushing module is used for pushing the recommendation list to the terminal corresponding to the user.
To achieve at least the above object, an embodiment of the present application further provides a server, including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor; so that the at least one processor can perform the recommendation method described above.
To achieve at least the above object, the embodiments of the present application further provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-mentioned recommendation method.
According to the recommendation method provided by the embodiment of the application, under the condition that the user is determined to be online, a user tag built offline for the user is obtained from an offline database; performing label matching according to the user labels and the content labels which are stored in the offline database and are constructed offline for each content to be recommended, and generating a recommendation list for the user online according to the matching result; and pushing the recommendation list to a terminal corresponding to the user. In other words, the recommendation method is divided into an offline part and an online part, the user tag and the content tag are built in the offline part, and the online part performs tag matching and pushing of the recommendation list, so that the online calculation amount of the server is reduced, the calculation efficiency is improved, and the recommendation response speed is accelerated.
Drawings
FIG. 1 is a flow chart of a recommendation method mentioned in an embodiment of the present application;
FIG. 2 is a block diagram of the implementation of the proposed method mentioned in the embodiments of the present application;
FIG. 3 is a schematic diagram of the constitution of the application layers mentioned in the embodiments of the present application;
FIG. 4 is a schematic illustration of both online and offline processes mentioned in the embodiments of the present application;
FIG. 5 is a schematic structural view of a recommending apparatus according to an embodiment of the present application;
fig. 6 is a schematic structural diagram of a server mentioned in an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. However, as will be appreciated by those of ordinary skill in the art, in the various embodiments of the present application, numerous technical details have been set forth in order to provide a better understanding of the present application. However, the technical solutions claimed in the present application can be implemented without these technical details and with various changes and modifications based on the following embodiments. The following embodiments are divided for convenience of description, and should not be construed as limiting the specific implementation of the present application, and the embodiments may be mutually combined and referred to without contradiction.
An embodiment of the present application provides a recommendation method applied to a server, where the server may be a server in a CDN network. The present embodiment is to provide recommended content for a user in a content delivery network (Content Delivery Network, CDN) network, such as content that the user views after entering an electronic program guide (electrical program guide, EPG) page, content that the user views after finishing a movie, and the like. The flow chart of the recommendation method may refer to fig. 1, including:
step 101: under the condition that the user is determined to be online, acquiring a user label built offline for the user from an offline database;
step 102: performing label matching according to user labels and content labels which are stored in an offline database and are constructed offline for each content to be recommended, and generating a recommendation list for the user online according to a matching result;
step 103: and pushing the recommendation list to the terminal corresponding to the user.
The inventor of the self application finds that the reasons of low calculation efficiency and low recommendation response speed when the current personalized recommendation system faces mass data are as follows: most of online real-time recommendation is adopted, and the whole operation process is completed online, so that the whole process of a machine learning algorithm with extremely high calculation force demand can be executed online, the execution speed of the machine learning algorithm is low, and the defect of high calculation force demand can be inherited by the whole recommendation system, so that the online calculation pressure is high, and the recommendation response speed is low. According to the embodiment of the application, the recommendation method is divided into an offline part and an online part, the user tag and the content tag are built in the offline part, and the online part carries out tag matching and pushing of the recommendation list, so that the online calculation amount of the server is reduced, the calculation efficiency is improved, and the recommendation response speed is accelerated. The embodiment is suitable for an application environment with the reduced operand, and the calculation force is reduced and the user recommendation response speed is increased through the split of the offline part and the online part. The embodiment is also suitable for improving the application environment of user response, saving calculation power and improving the user recommendation response speed through the separation of the offline part and the online part, and the online part can only provide the matching of the user and the content to be recommended in one recommendation process, so that the calculation amount is extremely low and the response speed is extremely high.
In step 101, the server may obtain, from an offline database, a user tag built offline for the user, if it is determined that the user is online watching audio and video. The off-line database can store a plurality of user labels of users, wherein the plurality of users can comprise users who watch audio and video on line, so that after a certain user is on line, the server can acquire the user labels of the users from the off-line database. The off-line database may be a database in the server or a database outside the server. That is, the user tag may be built off-line and stored in a database preset in the server, or stored in a database preset outside the server. The user tags may be used to characterize feature data of different dimensions of the user, such as, for example, the user's identity features, viewing record features, viewing preference features, etc. The server can directly acquire the user label which is built offline for the user in the offline database, does not need to acquire the user data online, and builds the user label online based on the online acquired user characteristics, so that the server can quickly acquire the user label of the user.
In step 102, the server may perform tag matching according to the user tag and the content tag that is stored in the offline database and is configured offline for each content to be recommended, and generate a recommendation list for the user online according to the matching result. The server may store a plurality of content tags of the content to be recommended, where the content to be recommended may include objects such as video and audio that may be watched or listened by the user. The content tag can be used for representing characteristic data of the content in different dimensions, the content tag can be obtained by the server through offline construction according to the characteristic data of the content to be recommended, and the content characteristics can comprise one or any combination of the following components: director of the content, actors in the content, type of content, score of the content, viewing population of the content, viewing time period of the content.
In one embodiment, tag matching may be understood as: the server carries out regression classification on the user label of the user and the content label of each content to be recommended stored in the offline database through a logistic regression (logistic regression, LR) technology, obtains the satisfaction degree value of the user on each content to be recommended, and generates a recommendation list for the user according to the satisfaction degree value of the user on each content to be recommended. For example, ranking the contents to be recommended from high to low according to the satisfaction value of the user, and taking the contents to be recommended ranked as the top N contents to be recommended as a recommendation list for the user. The value of N may be set according to actual needs, for example, may be 6, 7, 8, etc., however, this embodiment is not limited specifically.
In step 103, the server may push a recommendation list to a terminal corresponding to the user. The terminal corresponding to the user can be a terminal attached to the user on line, and the terminal can be a mobile phone, a television, a tablet personal computer and the like. For example, after a user turns on a television networking to watch, the server may send a recommendation list generated for the user to the television that the user watches, so that the user may watch the content recommended by the server.
In one embodiment, the user tag mentioned in step 101 is built offline by: performing offline feature engineering construction on the identity data of the user and the watching record data of the user to obtain a user tag; wherein the offline feature engineering configuration comprises at least one or any combination of the following: offline feature subdivision, offline feature mining, offline feature combining. Through offline feature subdivision, offline feature mining, offline feature combination and the like, accurate microscopic and comprehensive user labels are obtained, so that the accuracy of online matching is improved, and the accuracy of recommendation is improved.
Wherein the identity data of the user may include, but is not limited to, one or any combination of the following: the geographic location of the user, the viewing device of the user, the registration identity information of the user, etc. The user's viewing record data may include, but is not limited to, one or any combination of the following: user's search record, user's click record, user's subscription record, etc.
In one embodiment, the offline feature engineering construction is performed on the identity data of the user and the viewing record data of the user to obtain a user tag, which may include: and performing offline feature engineering construction on the identity data of the user and the watching record data of the user by adopting an XGBOOST algorithm so as to obtain an accurate and microscopic user tag. The set of user tags of a plurality of users may form a user feature model, that is, in this embodiment, the user tags may be abstracted from the identity data of the users and the viewing record data of the users, and the combination of the plurality of user tags constructs the user feature model.
In one embodiment, before the offline feature engineering construction is performed on the identity data of the user and the viewing record data of the user, the method may further include: performing offline preprocessing on the identity data of the user and the viewing record data of the user, wherein the offline preprocessing may include: 1) And integrating the watching record data of the user and the identity data of the user, so that the watching record data of the user is matched with the identity data of the user to obtain integrated data containing the watching record data of the user and the identity data of the user, and the integrated data can be supplied to an algorithm system for training. 2) The method has the advantages that the watching record data of the user and the identity data of the user are classified and cleaned, so that the watching record data of the user and the identity data of the user are in a data format which can be provided for subsequent processing, the situations of error leakage, vacancy, error and the like of the data are avoided, and the situations that the subsequent machine learning algorithm learning process is polluted or the operation data is polluted, so that model training errors are caused, model training results are wrong, recommended tasks cannot be executed, recommended task result errors are executed and the like are avoided.
In one embodiment, the content tags mentioned in step 102 are built offline by: performing offline feature engineering construction on content data of the content to be recommended and viewing record data of the content to be recommended to obtain a content tag; wherein the offline feature engineering configuration comprises at least one or any combination of the following: offline feature subdivision, offline feature mining, offline feature combining. Through offline feature subdivision, offline feature mining, offline feature combination and the like, accurate microscopic and comprehensive content labels are obtained, so that online matching accuracy is improved, and recommendation accuracy is improved.
Wherein the content data of the content to be recommended includes, but is not limited to, one of the following or any combination thereof: director of the content to be recommended, actors in the content to be recommended, type of the content to be recommended, score of the content to be recommended. The viewing record data for which the content to be recommended is viewed includes, but is not limited to, one of the following or any combination thereof: a viewing group (e.g., a child group, a office group, an elderly group, etc.), a period of time for which the content to be recommended is viewed, a place for which the content is viewed, and the like.
In one embodiment, the offline feature engineering construction of the content data of the content to be recommended and the viewing record data of the content to be recommended to be viewed may include: and performing offline feature engineering construction on the content data of the content to be recommended and the watching record data of the watched content to be recommended by adopting an XGBOOST algorithm so as to obtain a precise and fine microscopic content label. The set of content tags of the plurality of content to be recommended may form a content feature model, that is, in this embodiment, the content tags may be abstracted from the content data of the content to be recommended and the viewing record data in which the content to be recommended is viewed, and a combination of the plurality of content tags is constructed into the content feature model.
In one embodiment, before the offline feature engineering construction is performed on the content data of the content to be recommended and the viewing record data of the content to be recommended to be viewed, the method may further include: the method comprises the steps of classifying and cleaning content data of the content to be recommended and watching record data of the content to be recommended, so that the content data and the watching record data of the content to be recommended can be in a data format provided for subsequent processing, the situations of error leakage, vacancy, error and the like of the data are avoided, the situation that a subsequent machine learning algorithm learning process is polluted or operation data is polluted, model training errors are caused, model training results are wrong, a recommending task cannot be executed, and recommending task result errors are executed are avoided.
In one embodiment, in the case of determining that the user is online, the server may further acquire user data of the user and content data of content watched by the user online; after the step 103 sends the recommendation list to the terminal corresponding to the user, the method further includes: according to the online acquired user data, offline updating is carried out on the user tag; and updating the content label offline according to the content data of the content watched by the user online. It can be appreciated that the user data may not be constant, so in this embodiment, the user tag is updated offline according to the online acquired user data, which is beneficial to saving computing power through offline updating while obtaining the latest user tag that best meets the current characteristics of the user. Similarly, there is data that may change in the content data of the content, for example, the score of the content may change with time, so in this embodiment, the content tag is updated offline according to the content data obtained online, which is favorable for saving computing power through offline updating while obtaining the latest content tag that best meets the current content characteristics.
The offline build or offline update described above can be understood as: the construction or updating performed without a strong connection between the user and the server can also be understood as: the construction or updating of a server without networking can also be understood as: the server, after obtaining the relevant data, performs continuous work of non-instant response (i.e. without instant response to user operation). Under the condition of constructing or updating the user tag offline, the related data obtained by the server comprises the identity data of the user and the watching record data of the user; in the case of constructing or updating a content tag offline, the related data obtained by the server includes content data of the content and viewing record data in which the content is viewed. The following description will take the server offline construction or offline update of the user tag as an example:
for example, in the process that a user watches television online today, a server acquires identity data of the user and watching data of the user online, and then the server can construct or update a user tag offline according to the identity data and the watching record data acquired online. The continuous operation of the non-instant response by the server can be: after the user turns off the television, the server still continuously performs offline construction or updating of the user tag according to the user's identity data and viewing record data. It will be appreciated that the offline construction of the user tag is performed when the user has not been previously user-tag constructed by the server, i.e. the user tag of the user does not yet exist in the offline database, and the offline update of the user tag is performed when the user has been previously user-tag constructed by the server, i.e. the user tag of the user already exists in the offline database.
Assuming that the time period of watching television by the user is between 8 and 10 points in the evening, after the server acquires the identity data and the watching data of the user between 8 and 10 points, the server can start to perform offline construction or offline update of the user tag at any later time point. The duration of the offline build or offline update may last until the point in time when the user turns on the television next time, as the user typically has a longer time, such as 1 day, between turning on the television twice, and thus the server has a longer time to perform the offline build or offline update. Because the online construction or update of the tag generally requires a relatively high time, for example, the time required for the construction or update is relatively high, and thus the cost of the required server is relatively high, in this embodiment, the server may have a relatively long time for offline construction or offline update, so that the method can be completed by using a relatively low-cost server, that is, the method in this embodiment may further achieve the effect of reducing the cost.
In this embodiment, each time new content and new user are generated, the update of the user tag and the update of the content tag can be triggered, so that the user feature model and the content feature model can be trained in real time.
In one embodiment, the server may begin to acquire user data of the user and content data of content that the user views online upon determining that the user is online. In another embodiment, the server may begin to acquire the user data of the user and the content data of the content watched by the user online after pushing the recommendation list to the user, so as to provide enough calculation space for pushing the recommendation list online, thereby further improving recommendation efficiency.
In an embodiment, in step 103, the terminal is in a target scene, where the real-time requirement of the target scene on recommendation is lower than a preset requirement, the preset requirement may be set according to an actual requirement, that is, the real-time requirement of the user on recommendation in the target scene is lower, for example, the target scene may be a scene with the real-time requirement on recommendation lower than that of the internet scene, and it may be understood that the requirement of the short video application of the internet scene on recommendation real-time is generally higher. By executing the recommendation method in the embodiment of the application in the target scene, the recommendation efficiency is improved in the target scene, and the requirement of a user on instantaneity in the target scene is not influenced.
In one embodiment, the target scene is a scene of the terminal after entering the EPG page, the terminal may be an interactive internet protocol television (Internet Protocol TV, IPTV), and indexing and navigation of various services provided by the IPTV are all completed through the EPG system. Thus, the target scene may be understood as a scene with IPTV, such as a living room scene. In this embodiment, it is considered that in the living room scenario, the requirement for the recommended update speed is low, which can be understood as that the requirement for real-time recommendation is low, and the requirement for the recommended response speed is high. Therefore, for recommendation in living room scenes, the offline and online part splitting is beneficial to saving calculation force, and the recommendation response speed is accelerated under the condition that the requirement on the recommendation update speed is not influenced. For example, when a user opens IPTV in a living room, a user tag used when recommending a user is a user tag of the user stored in an offline database, and when a user opens IPTV in a living room, a user tag used when recommending a user is a user tag updated offline by user data of the user acquired online by monday stored in an offline database. That is, each time a user is online, the user label can be directly obtained from the constructed or updated user label stored in the offline database without the need of performing a long-time online user label construction process, so that the online recommendation efficiency is greatly improved.
In one embodiment, an architecture diagram implementing the recommendation method may refer to fig. 2, including: the system comprises a data layer, a data preprocessing layer, an application layer and a main control layer. The above layers may be distributed in the whole network of the CDN main network, have acquisition rights to all user data of the CDN, have acquisition rights to all content data in the CDN, and may all run in the CDN main server except the data layer, and the data layer may run in each packet server of the CDN. The functional architecture of these four levels will be described below:
the data layer is primarily responsible for acquiring various types of data including, but not limited to, user data and content data. This layer may exist in the CDN system as one entity network element (including but not limited to an entity server, a container of the fabric, etc.). The data layer comprises: the system comprises an identity data acquisition module of a user, a watching record data acquisition module of the user and a content data acquisition module.
The user identity data acquisition module is used for acquiring the identity information of the user including but not limited to the geographic position of the user, the viewing equipment of the user, the registered identity information of the user and the like, and constructing a user identity tag by taking the user identity information as a core.
And the viewing record data acquisition module is used for acquiring the behavior records of the user, including but not limited to search records of the user, click records of the user, subscription records of the user and the like, so as to construct the viewing preference label of the user for the core. The user tags mentioned in the above embodiments may include a user identity tag and a user's viewing preference tag.
The content data acquisition module is used for acquiring content related information including but not limited to a content provider, a content director, a content actor, a content classification, a content score and the like, and constructing a content label by taking the content related information as a core.
The data preprocessing layer is mainly responsible for classifying and cleaning various data acquired by the data layer, so that the data is in a data format which can be provided for processing by the application layer, the situations of error leakage, vacancy, error and the like of the data are avoided, and the situations that the learning process of a machine learning algorithm of the application layer is polluted or operation data is polluted, so that model training errors are caused, model training results are wrong, recommended tasks cannot be executed, recommended task result errors are executed and the like are avoided. This layer exists in the CDN system as one entity network element (including but not limited to an entity server, a container of the container, etc.). The data preprocessing layer comprises the following steps: the system comprises a user data preprocessing module and a content data preprocessing module.
The user data preprocessing module mainly comprises two functions. 1) And integrating the identity data of the user and the watching record data of the user, so that the watching record data of the user is matched with the identity data to obtain an integrated data supply algorithm system containing the identity data of the user and the watching record data for training. 2) The watching record and the identity data of the user are classified and cleaned, so that the watching record and the identity data are in a data format which can be provided for processing of an application layer, the situations of error leakage, vacancy, error and the like of the data are avoided, and the situations that a machine learning algorithm learning process of the application layer is polluted or operation data is polluted, so that model training errors are caused, model training results are wrong, recommended tasks cannot be executed, recommended task result errors are executed and the like are avoided.
The content data preprocessing module is mainly used for classifying and cleaning the content data acquired by the content data acquisition module, so that the content data can be provided with a data format for processing by an application layer, the situations of error leakage, vacancy, error and the like of the data are avoided, the situation that the learning process of a machine learning algorithm of the application layer is polluted or operation data is polluted, model training errors are caused, model training results are wrong, recommended tasks cannot be executed, and recommended task results are wrong are executed.
The application layer mainly comprises a label construction module for constructing labels; the recommended content generation module is used for tag matching and generation of a recommended result.
In one embodiment, the layering of the application layer may be as shown in fig. 3, including: the system comprises an original data layer, a mining feature combination layer, a prediction scoring layer and an output layer. In fig. 3, the flow before LR tag matching is an offline flow, and the flow after LR tag matching is an online flow. The various levels in fig. 3 are described below:
and the original data layer is mainly used for acquiring related data from the user data preprocessing module and the content data preprocessing module and carrying out characteristic engineering construction so as to obtain a user tag and a content tag. The original data layer mainly comprises three main data: USER identification data (denoted as USER MAP) USER viewing record data (denoted as LINK MAP) and content data ITEM MAP of the content to be recommended. The USER MAP is mainly obtained from identity data in the USER data preprocessing module, including but not limited to the geographic location of the USER, viewing equipment of the USER, registered identity information of the USER and other identity information of the USER, and the USER identity tag is initially built and completed by taking the USER MAP as a core. The LINK MAP is mainly obtained from viewing record data in the user data preprocessing module, including but not limited to search records of users, click records of users, subscription records of users, and the like, so as to construct a viewing preference tag of the users for the core, and enable the viewing preference tag of the users to be associated with the user identity tag. The ITEM MAP is mainly obtained from content data in a content data preprocessing module, and comprises content related information such as a content provider, a content director, a content actor, a content classification, a content grading and the like, and the content label is primarily built by taking the content related information as a core.
Digging a characteristic combination layer: the layer mainly utilizes the USER MAP, the LINK MAP and the ITEM MAP to carry out offline feature engineering construction through an XGBOOST algorithm, such as feature subdivision, feature mining and automatic feature combination, and the accurate and fine microcosmic new USER tags and content tags are constructed. And (3) carrying out automatic information mining and feature combination on identity data USER MAP and viewing record data LINK MAP of a single USER to construct an accurate microscopic USER tag of the USER. Through XGBOOST feature subdivision, feature mining and automatic feature combination of content data of the content to be recommended and viewing record data of the content to be recommended, an accurate microscopic content tag is constructed.
And (3) predicting the evaluation layer, when a target user watches online, acquiring a user tag constructed for the target user in an offline database, carrying out regression classification (namely tag matching) on the accurate microscopic user tag of the target user and the accurate microscopic content tag of each content to be recommended in the database through an LR technology, predicting the satisfaction degree value of the target user for each content to be recommended, and sequencing the satisfaction degree value of each content to be recommended through the user to obtain a recommendation list for the target user.
And the output layer is used for pushing the recommendation list to a client used by a user.
And (3) a main control layer: the main control layer mainly comprises a main control module which is used for controlling the operation of the recommendation algorithm, issuing the total quantity of certain hot content, uniformly regulating and controlling black and white lists of certain users and the like through the main control layer. This layer exists in the CDN system as one entity network element (including but not limited to an entity server, a container of the container, etc.).
In one embodiment, in the application layer, the above-mentioned two processes are mainly divided into online and offline, and the layer exists as one entity network element (including but not limited to an entity server, a DOCKER container, etc.) in the CDN system. The flow chart is shown in fig. 4, and the following is a simple description:
when the server detects that the user is online, the online process is executed: generating a recommendation list based on the user; the recommendation list is pushed to the user. Specifically, a server acquires a user tag built offline for the user from an offline database; and performing tag matching according to the user tags and content tags of the contents to be recommended stored in the offline database, generating a recommendation list for the user on line according to the matching result, and pushing the recommendation list to the user.
When the server detects that the user is online, the online process further performs: acquiring identity data of a user, viewing record data watched by the user on line and content data of content watched by the user on line;
the offline flow is executed: importing data into an offline database; updating the user tag; the content tag is updated. The data imported into the offline database comprises user identity data acquired in an online process, viewing record data watched by the user online and content data of content watched by the user online; the user tag can be updated according to the identity data of the user and the viewing record data of the online viewing of the user, which are imported into the offline database, and the content tag can be updated according to the content data of the online viewing of the user.
In the embodiment, a recommendation algorithm system is built in an offline and online mode, and a recommendation list is generated online by building a user tag and a content tag, so that the balance of operation speed and calculation resources is achieved. By arranging machine learning for constructing the user tag and the content tag to an offline part, the online computing pressure is reduced, the recommendation list generation rate is improved, and the user experience is improved. By arranging the recommendation list generation to the online part, the recommendation content is updated in real time, so that the refresh speed of the recommendation content is ensured, and the user experience is improved. The machine learning is continuously self-iterated in the running process of the system, and the feature extraction capability is continuously self-reinforced, so that the recommendation success rate can be improved.
It should be noted that, the foregoing examples in the embodiments of the present application are all illustrative for easy understanding, and do not limit the technical solution of the present invention.
The above steps of the methods are divided, for clarity of description, and may be combined into one step or split into multiple steps when implemented, so long as they include the same logic relationship, and they are all within the protection scope of this patent; it is within the scope of this patent to add insignificant modifications to the algorithm or flow or introduce insignificant designs, but not to alter the core design of its algorithm and flow.
One embodiment of the present application provides a recommendation device, as shown in fig. 5, including: the offline user tag obtaining module 501 is configured to obtain, from an offline database, a user tag configured offline for a user, in case that it is determined that the user is online; the online matching module 502 is configured to perform tag matching according to the user tag and the content tag that is stored in the offline database and is built offline for each content to be recommended, and generate a recommendation list for the user online according to the matching result; and the online pushing module 503 is configured to push the recommendation list to a terminal corresponding to the user.
It should be noted that, each module related to this embodiment is a logic module, and in practical application, one logic unit may be one physical unit, or may be a part of one physical unit, or may be implemented by a combination of multiple physical units. In addition, in order to highlight the innovative part of the present invention, units less closely related to solving the technical problem presented by the present invention are not introduced in the present embodiment, but it does not indicate that other units are not present in the present embodiment.
It is to be noted that this embodiment is an embodiment of the apparatus corresponding to the above-described method embodiment, and this embodiment may be implemented in cooperation with the above-described method embodiment. The related technical details and technical effects mentioned in the above method embodiments are still valid in this embodiment, and in order to reduce repetition, they are not repeated here. Accordingly, the related technical details mentioned in the present embodiment can also be applied in the above-described method embodiments.
One embodiment of the present application provides a server, as shown in fig. 6, comprising: at least one processor 601; and a memory 602 communicatively coupled to the at least one processor 601; the memory 602 stores instructions executable by the at least one processor 601, and the instructions are executed by the at least one processor 601 to enable the at least one processor 601 to perform the recommendation method.
Where the memory 602 and the processor 601 are connected by a bus, the bus may comprise any number of interconnected buses and bridges, the buses connecting the various circuits of the one or more processors 601 and the memory 602. The bus may also connect various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art, and therefore, will not be described any further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or may be a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. The data processed by the processor 601 is transmitted over a wireless medium via an antenna, which further receives the data and transmits the data to the processor 601.
The processor 601 is responsible for managing the bus and general processing and may also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. And memory 602 may be used to store data used by processor 601 in performing operations.
Embodiments of the present application also provide a computer-readable storage medium storing a computer program. The computer program implements the above-described method embodiments when executed by a processor.
That is, it will be understood by those skilled in the art that all or part of the steps in implementing the methods of the embodiments described above may be implemented by a program stored in a storage medium, where the program includes several instructions for causing a device (which may be a single-chip microcomputer, a chip or the like) or a processor (processor) to perform all or part of the steps in the methods of the embodiments described herein. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples of carrying out the invention and that various changes in form and details may be made therein without departing from the spirit and scope of the invention.
Claims (11)
1. A recommendation method, comprising:
under the condition that the user is determined to be online, acquiring a user tag built offline for the user from an offline database;
performing tag matching according to the user tag and the content tag which is stored in the offline database and is constructed offline for each content to be recommended, and generating a recommendation list for the user online according to a matching result;
and pushing the recommendation list to the terminal corresponding to the user.
2. The recommendation method according to claim 1, wherein in case it is determined that the user is online, the method further comprises:
acquiring user data of the user and content data of content watched by the user online;
after the recommendation list is sent to the terminal corresponding to the user, the method further comprises the following steps:
according to the user data obtained online, the user label is updated offline;
and updating the content label offline according to the content data of the content watched by the user online.
3. The recommendation method according to claim 1, wherein said user tag is built offline by:
performing offline feature engineering construction on the identity data of the user and the watching record data of the user to obtain the user tag; wherein the offline feature engineering configuration comprises at least one or any combination of the following: offline feature subdivision, offline feature mining, offline feature combining.
4. The recommendation method according to claim 1, wherein said content tags are built offline by:
performing offline feature engineering construction on the content data of the content to be recommended and the watching record data of the watched content to be recommended to obtain the content tag; wherein the offline feature engineering configuration comprises at least one or any combination of the following: offline feature subdivision, offline feature mining, offline feature combining.
5. The recommendation method according to claim 2, wherein the user data acquired online includes identity data of the user and viewing record data that the user views online.
6. The recommendation method according to claim 2, wherein the content data of the content viewed online by the user comprises any combination of:
the director of the content, the actors in the content, the type of the content, the score of the content, the viewing population of the content, the viewing time period of the content.
7. The recommendation method according to any one of claims 1 to 6, wherein the terminal is in a target scene, and the real-time requirement of the target scene on recommendation is lower than a preset requirement.
8. The recommendation method of claim 7, wherein the target scene is a scene after the terminal enters an electronic program guide EPG page.
9. A recommendation device, comprising:
the off-line user tag acquisition module is used for acquiring the user tag constructed offline for the user from an off-line database under the condition that the user is determined to be on-line;
the online matching module is used for carrying out tag matching according to the user tag and the content tag which is stored in the offline database and is constructed for each content to be recommended offline, and generating a recommendation list for the user online according to a matching result;
and the online pushing module is used for pushing the recommendation list to the terminal corresponding to the user.
10. A server, comprising: at least one processor; the method comprises the steps of,
a memory communicatively coupled to the at least one processor; wherein, the liquid crystal display device comprises a liquid crystal display device,
the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor; to enable the at least one processor to perform the recommendation method according to any one of claims 1 to 8.
11. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the recommendation method according to any one of claims 1 to 8.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302296.9A CN116070014A (en) | 2021-11-04 | 2021-11-04 | Recommendation method, recommendation device, server and computer readable storage medium |
| PCT/CN2022/128878 WO2023078226A1 (en) | 2021-11-04 | 2022-11-01 | Recommendation method and apparatus, server and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111302296.9A CN116070014A (en) | 2021-11-04 | 2021-11-04 | Recommendation method, recommendation device, server and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116070014A true CN116070014A (en) | 2023-05-05 |
Family
ID=86175726
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111302296.9A Pending CN116070014A (en) | 2021-11-04 | 2021-11-04 | Recommendation method, recommendation device, server and computer readable storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN116070014A (en) |
| WO (1) | WO2023078226A1 (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7533399B2 (en) * | 2004-12-02 | 2009-05-12 | Panasonic Corporation | Programming guide content collection and recommendation system for viewing on a portable device |
| CN108734510B (en) * | 2018-04-23 | 2021-09-14 | 微梦创科网络科技（中国）有限公司 | Advertisement recommendation method and system based on attribute matching |
| CN110533515A (en) * | 2019-09-04 | 2019-12-03 | 深圳创新奇智科技有限公司 | A kind of the electric business personalized recommendation method and device of high low latency of handling up |
| CN112597395A (en) * | 2020-12-28 | 2021-04-02 | 上海众源网络有限公司 | Object recommendation method, device, equipment and storage medium |
-
2021
- 2021-11-04 CN CN202111302296.9A patent/CN116070014A/en active Pending
-
2022
- 2022-11-01 WO PCT/CN2022/128878 patent/WO2023078226A1/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| WO2023078226A1 (en) | 2023-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111708901B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
| US20180077250A1 (en) | Method and System for Processing Data Used By Creative Users to Create Media Content | |
| CN108595493B (en) | Media content pushing method and device, storage medium and electronic device | |
| CN109688479B (en) | Bullet screen display method, bullet screen display device and bullet screen display server | |
| CN103986787A (en) | Method and system for automatically obtaining updating of video | |
| JP2015201157A (en) | Dynamic content recommendation system using social network data | |
| CN110139162A (en) | The sharing method and device of media content, storage medium, electronic device | |
| US20190034997A1 (en) | System and method of providing object for service of service provider | |
| CN112822526A (en) | Video recommendation method, server and readable storage medium | |
| CN114461871B (en) | Recommendation model training method, object recommendation device and storage medium | |
| CN111523050B (en) | Content recommendation method, server and storage medium | |
| CN113836406A (en) | Information flow recommendation method and device | |
| CN112287799A (en) | Video processing method and device based on artificial intelligence and electronic equipment | |
| KR20210097432A (en) | Apparatus and method for recommending contents | |
| US20220360836A1 (en) | User defined content summary channel | |
| CN116070014A (en) | Recommendation method, recommendation device, server and computer readable storage medium | |
| US20210365800A1 (en) | Computing apparatus and operating method thereof | |
| CN112269942B (en) | Method, device and system for recommending object and electronic equipment | |
| CN111859221B (en) | Project recommendation method and device | |
| CN116361508A (en) | Video recommendation method and related equipment | |
| CN115244941A (en) | User interface for improved video packaging | |
| CN114969493A (en) | Content recommendation method and related device | |
| CN113886674A (en) | Resource recommendation method and device, electronic equipment and storage medium | |
| CN114615524B (en) | Training method of server and media asset recommendation network and media asset recommendation method | |
| CN109254812A (en) | A kind of method, system, terminal and server updating Shipping Options Page |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication |