Media personalized recommendation feed service system construction method and feed service system

Abstract

The invention relates to a construction method of a media personalized recommendation feed service system and the feed service system, and the method comprises the steps: obtaining recommendation data from a data layer through an access data access service, and constructing a preliminary recommendation candidate set; the recommendation candidate set is dynamically updated by accessing a real-time correction service, and the storage and access efficiency of the recommendation content is optimized by using a three-level cache mechanism; by accessing a recommendation intervention service and an AB test service, dynamically adjusting a content weight and a display sequence in the recommendation candidate set according to a preset specification; and sorting, filtering and outputting the optimized recommendation candidate set to a user through accessing a recommendation interface service, and providing personalized media recommendation content and outputting the personalized media recommendation content to an application layer. By caching the high-frequency candidate set data and optimizing the recommendation strategy, the response speed of the system and the user experience are improved, the correlation and diversity of the media recommendation content are ensured, and meanwhile, the personalized requirement and the forward guidance standard of the user are met.

Description

Technical Field

[0001] The present invention relates to the technical field of personalized media recommendation, and in particular to a method for constructing a personalized media recommendation feed service system and a feed service system. Background Art

[0002] In the existing technology, personalized media recommendation systems have been widely used in various content platforms, such as news, videos, social media, etc. These systems usually generate personalized media recommendations by analyzing user behavior data (such as browsing history, click records, likes and comments, etc.) and combining the feature information of the content. However, the existing recommendation systems still have some shortcomings in processing efficiency, real-time performance and recommendation accuracy.

[0003] First, the existing system has bottlenecks in data processing efficiency. Due to the large amount and complexity of user behavior data, traditional batch processing methods are difficult to meet the needs of real-time recommendations, resulting in a decline in user experience. Secondly, the real-time nature of media recommendation content is insufficient. Existing systems are often unable to update user preferences and content status in a timely manner, resulting in lagging media recommendation content and failing to accurately reflect users' current interests. In addition, the flexibility and accuracy of recommendation strategies in existing systems need to be improved. Many systems lack the ability to dynamically adjust recommendation strategies and are unable to optimize based on real-time data and content specifications, resulting in insufficient personalization and accuracy in recommendation results.

[0004] In order to overcome the above shortcomings, an efficient, real-time and flexible personalized recommendation system is needed to improve the relevance of media recommendation content and user satisfaction. Summary of the invention

[0005] The purpose of the present invention is to provide a method for constructing a media personalized recommendation feed service system and a feed service system to solve the problems of low data processing efficiency, insufficient real-time performance and poor user experience in the prior art.

[0006] To achieve one of the above-mentioned purposes, an embodiment of the present invention provides a method for constructing a media personalized recommendation feed service system, which is characterized by comprising: Obtain recommendation data from the data layer by accessing the data access service and construct a preliminary recommendation candidate set; Dynamically update the recommendation candidate set by accessing the real-time correction service, and optimize the storage and access efficiency of the recommended content by using a three-level cache mechanism; By accessing the recommendation intervention service and AB testing service, the content weight and display order in the recommendation candidate set are dynamically adjusted according to the preset specifications; Sort, filter, and output the optimized recommended candidate set to the user through the access to the recommendation interface service, provide personalized media recommendation content, and output it to the application layer.

[0007] As a further improvement of an embodiment of the present invention, the method further includes analyzing the data in the recommended candidate set based on user behavior data, content popularity, and user characteristics through accessing the recommendation cache service, and caching the recommended candidate set data determined to be high-frequency.

[0008] As a further improvement of an embodiment of the present invention, the method further includes archiving the recommended candidate set by accessing the recommendation archiving service to implement the traceability query function of the recommendation data.

[0009] As a further improvement of an embodiment of the present invention, the method further includes that the three-level cache mechanism includes hierarchical storage through mongodb, elasticsearch, and redis; The mongodb is used to store the recommendation data, the elasticsearch is used to distribute and store the recommended candidate set; the redis is used to store the recommended candidate set determined to be high-frequency; Through the scheduled task and active user prediction, preload the recommended candidate set from elasticsearch to redis to achieve the balance between efficient query and fast access.

[0010] To achieve one of the above-mentioned invention purposes, an embodiment of the present invention provides a feed service system, which includes a data layer, a data processing layer, an interface service layer, and an application layer; Collect media data and user behavior data through the data layer and store them in the database; when receiving user request parameters, obtain the recommended candidate set from the database, and send it to the interface service layer after updating the recommended candidate set through the data processing layer; optimize, screen, and sort the recommended candidate set through the interface service layer, and finally output it to the application layer.

[0011] As a further improvement of an embodiment of the present invention, the system further includes that the data processing layer includes a recommended data access service module, a recommended real-time correction service module, a recommended cache service module, and a recommended archiving service module; the interface service layer includes a recommended interface service module, a recommended intervention service module, and a recommended AB test service module; The recommended data access service module is used to receive and process media data and user behavior data to provide a recommended candidate set for subsequent recommendations; The recommended real-time correction service module is used to update the recommended candidate set in real time, adjust the status of the read content, and optimize the timeliness and accuracy of the candidate set; The recommendation cache service module is used to store the recommendation candidate sets of high-frequency users, improve the real-time and efficiency of data access, and regularly update and clean up the cache; The recommendation archiving service module is used to archive and store the recommendation data to ensure the integrity and traceability of historical data for analysis and optimization; The recommendation interface service module is used to receive user request parameters, obtain a recommendation candidate set, and output the final media recommendation content after sorting and filtering the data; The recommendation intervention service module is used to dynamically adjust the weight and display order of media recommendation content according to preset media content specifications to ensure that the media recommendation content meets the personalized and positive guidance standards; The recommendation AB testing service module is used to group users to test different recommendation strategies, ensuring that each user uses the same strategy during the test to evaluate the effect.

[0012] As a further improvement of an embodiment of the present invention, the system further includes: the feed service system adopts a linkage three-level cache mechanism, including: The recommendation data access service module stores the recommendation data through the mongodb, the recommendation real-time correction service module stores the recommendation candidate set in a distributed manner through the elasticsearch; and the recommendation cache service module stores the recommendation candidate set determined to be high frequency through the redis.

[0013] As a further improvement of an embodiment of the present invention, the system further includes: the recommendation cache service module periodically queries active users within a preset time, and stores the recommendation candidate sets of the active users in redis; The redis uses the sorted set zset for storage and uses the user's unique identifier as the cache access key; The recommendation interface service module obtains a recommendation candidate set through the user unique identifier; The media that the user has read and added to the blacklist are cached in redis through the recommendation cache service module, and the set collection is used to cache according to the user's unique identifier, and the media that has been read and added to the blacklist are filtered before returning the recommendation candidate set to the user.

[0014] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention further provides an electronic device, comprising a memory and a processor, characterized in that the memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the method for constructing the above-mentioned media personalized recommendation feed service system are implemented.

[0015] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention further provides a storage medium, wherein the storage medium stores a computer program, and is characterized in that when the computer program is executed by a processor, the steps in the method for constructing the above-mentioned media personalized recommendation feed service system are implemented.

[0016] Compared with the prior art, the present invention provides a method for constructing a personalized media recommendation feed service system and a feed service system, which receives and standardizes multiple data sources, updates the recommendation candidate set in real time, uses AB testing to group users, and dynamically adjusts the weight and order of media recommendation content, thereby ensuring the high quality, personalization and real-time nature of media recommendation content. By caching high-frequency candidate set data and optimizing recommendation strategies, the system's response speed and user experience are improved, ensuring the relevance and diversity of media recommendation content, while meeting users' personalized needs and positive guidance standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is an overall flow chart of the method for constructing the media personalized recommendation feed service system of the present invention.

[0018] Figure 2 It is a schematic diagram of the architecture of the feed service system of the present invention.

[0019] Figure 3 It is a data flow diagram between microservices of the feed service system of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be described in detail below in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by a person skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0021] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0022] In the first embodiment of the present invention, the present invention provides a method for constructing a media personalized recommendation feed service system, such as Figure 1 As shown, the method includes, S1: Obtain recommendation data from the data layer by accessing the data access service and construct a preliminary recommendation candidate set; S2: Dynamically update the recommendation candidate set by accessing the real-time correction service, and optimize the storage and access efficiency of the recommended content by using a three-level cache mechanism; S3: By accessing the recommendation intervention service and AB testing service, the content weight and display order in the recommendation candidate set are dynamically adjusted according to the preset specifications; S4: sort, filter and output the optimized recommendation candidate set to the user through the access recommendation interface service, provide personalized media recommendation content and output it to the application layer.

[0023] In a specific embodiment of the present invention, by accessing the recommendation cache service, the data in the recommendation candidate set is analyzed based on user behavior data, content popularity and user characteristics, and the recommendation candidate set data determined to be high frequency is cached.

[0024] It should be noted that there are three types of external data sources in the data layer accessed through the recommendation data access service. The first type is media data, including published media data and recommendation intervention data. The second type is user behavior data, which is sent to the feed service system through the data collection service. The third type is recommendation data, which is the recommendation data and cold start data calculated by the recommendation engine module through the algorithm.

[0025] In a specific implementation of the present invention, the recommendation candidate set is archived by accessing a recommendation archiving service, thereby realizing a traceability query function of the recommendation data.

[0026] Furthermore, in order to ensure the query performance of the interface, it is necessary to archive the historical recommendation candidate sets regularly. To this end, the recommendation archiving service regularly cleans up the recommended candidate sets on elasticsearch and the keys on redis, archives these historical data to the object storage OSS, and deletes the storage of these data on redis and elasticsearch.

[0027] In a specific embodiment of the present invention, the three-level cache mechanism includes hierarchical storage through mongodb, elasticsearch and redis; The mongodb is used to store the recommendation data, the elasticsearch is used to store the recommendation candidate sets in a distributed manner; the redis is used to store the recommendation candidate sets determined to be high frequency; Through scheduled tasks and active user prediction, the recommendation candidate set is preloaded from elasticsearch to redis to achieve a balance between efficient query and fast access.

[0028] It should be noted that if Figure 3As shown in the figure, elasticsearch, mongodb and redis are used to implement a linked three-level cache. Through scheduled tasks and active user prediction mechanisms, the system predicts which users will frequently access recommended content in the next period of time based on user behavior and activity, and preloads the recommended candidate sets of these active users from Elasticsearch to Redis in advance. This mechanism ensures that when users access, the system can respond quickly and obtain the stored candidate set data from Redis, greatly improving query efficiency and reducing access latency. At the same time, Elasticsearch still maintains its role as a long-term storage to ensure the persistence and integrity of large-scale data. This approach achieves a balance between efficient query and fast access, reduces storage costs, and ensures the real-time and accuracy of recommendation services.

[0029] In the second embodiment of the present invention, the present invention provides a feed service system, such as Figure 2 As shown, the system includes: a data layer 100, a data processing layer 200, an interface service layer 300 and an application layer 400; Media data and user behavior data are collected through the data layer 100 and stored in a database; when user request parameters are received, a recommendation candidate set is obtained from the database, and the recommendation candidate set is updated through the data processing layer 200 and sent to the interface service layer 300; the recommendation candidate set is optimized, screened and sorted through the interface service layer 300, and finally output to the application layer 400.

[0030] Specifically, the data processing layer includes a recommended data access service module 201, a recommended real-time correction service module 202, a recommended cache service module 203, and a recommended archiving service module 204; the interface service layer 300 includes a recommended interface service module 301, a recommended intervention service module 302, and a recommended AB testing service module 303; The recommendation data access service module 201 is used to receive and process media data and user behavior data to form a recommendation candidate set for subsequent recommendation; The recommendation real-time correction service module 202 is used to update the recommendation candidate set in real time, adjust the read content status and optimize the real-time performance and accuracy of the candidate set; The recommendation cache service module 203 is used to store the recommendation candidate sets of high-frequency users, improve the real-time and efficiency of data access, and regularly update and clean up the cache; The recommendation archiving service module 204 is used to archive and store the recommendation data to ensure the integrity and traceability of historical data for analysis and optimization; The recommendation interface service module 301 is used to receive user request parameters, obtain a recommendation candidate set, and output the final media recommendation content after sorting and filtering the data; The recommendation intervention service module 302 is used to dynamically adjust the weight and display order of media recommendation content according to preset media content specifications to ensure that the media recommendation content meets the personalized and positive guidance standards; The recommendation AB testing service module 303 is used to group users and test different recommendation strategies to ensure that each user uses the same strategy during the test to evaluate the effect.

[0031] It should be noted that if Figure 2 As shown in the figure, the feed service system architecture is mainly divided into four layers: The first layer is the data layer 100, which receives data from the outside, including media data, user data, behavior data, and recommendation candidate sets from the recommendation engine. These four types of data are transmitted to the recommendation feed service according to the data format defined by the recommendation feed service; The second layer is the data processing layer 200, including the recommended data access service module 201, which distributes and stores different data pushed to the interface and performs data isolation storage. The recommended real-time correction service module 202 corrects the data status of the recommended candidate set. The recommended cache service module 203 regularly updates the candidate set data to the cache so that the data accessed by the recommended interface is the latest cache. The recommended archiving service module 204 archives and cleans up the data in the cache and distributed index to prevent the elasticsearch data from being too large, resulting in slow interface service queries; The third layer is the interface service layer 300, which includes the following three services: the recommendation interface service module 301 is mainly for various recommendation channels, provides the final recommendation result data, and receives real-time parameter information from users. For example, it can obtain the latest candidate set data sorting according to user hashId, geographic location, preferences, etc. The recommendation intervention service module 302 processes the candidate set data to be recommended to ensure that the recommended data meets positive energy and improves the client's end phase. The recommendation AB test service module 303 is mainly used to verify the effects of different recommendation strategies online, and facilitate the online adjustment of parameters of the recommendation strategy; The fourth layer is the application layer 400, which can be applied to different types of article type recommendations for different APPs, and can also adapt to recommendations for different channels. As long as data is passed in according to the specifications defined by the recommendation feed service, the recommendation interface service provided by the recommendation feed service can be obtained.

[0032] It should be noted that the recommendation interface service module 301 provides a recommendation interface data service, and returns the recommendation data of the corresponding channel according to the corresponding requested user hashId and channel Id. At the same time, the recommendation interface service module 301 needs to use the data pre-processed by the other 6 microservices as above to select the strategy.

[0033] It should be noted that in order to make the recommended content more in line with the characteristics of the media, it is necessary to intervene in the recommendation process data before, during and after the recommendation process through the recommendation intervention service module 302. Before the event, the data of the corresponding column can be controlled to allow the recommendation of the manuscript pool and which manuscripts cannot enter the recommendation pool; during the event, the recommendation weight of certain manuscripts can be specified to increase or decrease the recommendation opportunity of this manuscript; after the event, the recommendation interface layer can be added with recommendation pinning or recommendation blocking to make the recommendation result more controllable and ensure the safety of media recommendation content.

[0034] It should be noted that by recommending AB test service module 303 to switch AB schemes for different users, the corresponding algorithm processing can be adjusted in real time according to the performance of online users. The access ratio of each scheme is set in the AB service, and then bucket storage is performed according to the user hashId. When a user is assigned to scheme A, he will continue to access scheme A until the experiment of this AB scheme is cleared. The service also supports importing the corresponding scheme specified by the user.

[0035] It should be noted that if Figure 3 As shown, the feed service system has data access including scheduling distribution, recommendation engine and data collection services, and data output including mobile phone interface service and OSS object storage.

[0036] It should be noted that by splitting the recommendation feed service into multiple microservices and using Kubernetes for container orchestration, the microservices can run independently. By using redis cache, NoSQL database and distributed index database, the business code and storage status are completely separated. When encountering business peaks, the recommendation feed service can be quickly and elastically expanded to cope with user access during peak periods.

[0037] In a specific embodiment of the present invention, the feed service system adopts a linkage three-level cache mechanism, including: The recommendation data access service module 201 stores the recommendation data through the mongodb, the recommendation real-time correction service module 202 stores the recommendation candidate set through the elasticsearch distributed storage; the recommendation cache service module 203 stores the recommendation candidate set determined to be high frequency through the redis.

[0038] It should be noted that MongoDB is used to store detailed information about recommendation data, including user behavior data, metadata of media content, and historical recommendation records. As a NoSQL database, MongoDB can provide a flexible data model and efficient query performance, and is suitable for storing complex and frequently changing data. Elasticsearch is used to store recommendation candidate sets in a distributed manner. Through its powerful full-text search and distributed query capabilities, Elasticsearch can quickly retrieve and index recommendation candidate sets, ensuring efficient access to massive data. The metadata of recommendation candidate sets, content popularity indicators, user characteristics, and other information are all stored in Elasticsearch, facilitating fast search and dynamic updates. Redis is used to store recommendation candidate sets that are determined to be high-frequency, especially recommended content for currently active users. Redis is suitable for data access scenarios that require low latency and high throughput through its high-performance memory storage and fast data access capabilities. Redis sorts and stores candidate sets through its zset (ordered set) data structure, supports fast access by priority or time sorting, and ensures that the recommendation system can provide the best recommended content when responding to user requests instantly.

[0039] In a specific embodiment of the present invention, the recommendation cache service module 203 periodically queries active users within a preset time, and stores the recommendation candidate set of the active users in redis; the redis uses the sorted set zset for storage, and uses the user's unique identifier as the cache access key; The recommendation interface service module 301 obtains a recommendation candidate set through the user unique identifier; The media that the user has read and added to the blacklist are cached in redis through the recommendation cache service module 203, and the set collection is used to cache according to the user unique identifier, and the media that has been read and added to the blacklist are filtered before returning the recommendation candidate set to the user.

[0040] It should be noted that, through the recommendation cache service module 203, the system will periodically query active users within a preset time, so as to store the recommendation candidate sets of these users in the Redis cache in advance. In order to effectively manage and store this data, the system uses Redis sorted sets (zset) for storage, and uses each user's unique identifier hashId as the key value for cache access. In this way, the interface service only needs to use the user's hashId to quickly obtain the user's candidate set to be recommended, thereby improving the efficiency and real-time performance of data access.

[0041] Furthermore, the content that users have read and the articles that have been blacklisted will also be cached in Redis. For ease of management, these data will be cached by user using the Redis set structure. Before returning the recommendation results to the user, the system will filter based on these cached data to ensure the accuracy and relevance of the recommended content. In this way, not only the use of cache resources is optimized, but also the performance of the recommendation system and user experience are significantly improved.

[0042] In a third embodiment of the present invention, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the method for constructing the above-mentioned media personalized recommendation feed service system are implemented.

[0043] In a fourth embodiment of the present invention, the present invention provides a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps in the method for constructing the above-mentioned media personalized recommendation feed service system.

[0044] In summary, the present invention provides a method for constructing a personalized media recommendation feed service system and a feed service system, which receives and standardizes multiple data sources, updates the recommendation candidate set in real time, uses AB testing to group users, and dynamically adjusts the weight and order of media recommendation content, thereby ensuring the high quality, personalization, and real-time nature of media recommendation content. By caching high-frequency candidate set data and optimizing recommendation strategies, the system's response speed and user experience are improved, ensuring the relevance and diversity of media recommendation content, while meeting users' personalized needs and positive guidance standards.

[0045] It should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation mode may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0046] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the modules described above can refer to the corresponding process in the aforementioned method implementation, and will not be repeated here.

[0047] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present implementation scheme.

[0048] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0049] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0050] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned implementation modes, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned implementation modes, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various implementation modes of the present application.

Claims

1. A method for constructing a media personalized recommendation feed service system, characterized in that: include, Obtain recommendation data from the data layer by accessing the data access service and construct a preliminary recommendation candidate set; Dynamically update the recommendation candidate set by accessing the real-time correction service, and optimize the storage and access efficiency of the recommended content by using a three-level cache mechanism; By accessing the recommendation intervention service and AB testing service, the content weight and display order in the recommendation candidate set are dynamically adjusted according to preset specifications; By accessing the recommendation interface service, the optimized recommendation candidate set is sorted, filtered and output to the user, providing personalized media recommendation content and outputting it to the application layer.

2. The method for constructing a personalized media recommendation feed service system according to claim 1, characterized in that: Also includes, By accessing the recommendation cache service, the data in the recommendation candidate set is analyzed based on user behavior data, content popularity and user characteristics, and the recommendation candidate set data determined to be high-frequency is cached.

3. The method for constructing a personalized media recommendation feed service system according to claim 1, characterized in that: Also includes, The recommendation candidate set is archived by accessing the recommendation archiving service to realize the traceability query function of the recommendation data.

4. The method for constructing a personalized media recommendation feed service system according to claim 1, characterized in that: The three-level cache mechanism includes hierarchical storage through mongodb, elasticsearch and redis; The mongodb is used to store the recommendation data, the elasticsearch is used to store the recommendation candidate sets in a distributed manner; the redis is used to store the recommendation candidate sets determined to be high frequency; Through scheduled tasks and active user prediction, the recommendation candidate set is preloaded from elasticsearch to redis to achieve a balance between efficient query and fast access.

5. A feed service system, characterized in that: include, Data layer, data processing layer, interface service layer and application layer; Collect media data and user behavior data through the data layer and store them in a database; When receiving the user request parameters, obtaining the recommendation candidate set from the database, updating the recommendation candidate set through the data processing layer and sending it to the interface service layer; The recommendation candidate set is optimized, screened and sorted through the interface service layer, and finally output to the application layer.

6. The feed service system according to claim 5, characterized in that: The data processing layer includes a recommended data access service module, a recommended real-time correction service module, a recommended cache service module and a recommended archiving service module; the interface service layer includes a recommended interface service module, a recommended intervention service module and a recommended AB testing service module; The recommendation data access service module is used to receive and process media data and user behavior data to prepare a recommendation candidate set for subsequent recommendation; The recommendation real-time correction service module is used to update the recommendation candidate set in real time, adjust the status of the read content and optimize the real-time performance and accuracy of the candidate set; The recommendation cache service module is used to store the recommendation candidate sets of high-frequency users, improve the real-time and efficiency of data access, and regularly update and clean up the cache; The recommendation archiving service module is used to archive and store the recommendation data to ensure the integrity and traceability of historical data for analysis and optimization; The recommendation interface service module is used to receive user request parameters, obtain a recommendation candidate set, and output the final media recommendation content after sorting and filtering the data; The recommendation intervention service module is used to dynamically adjust the weight and display order of media recommendation content according to preset media content specifications to ensure that the media recommendation content meets the personalized and positive guidance standards; The recommendation AB testing service module is used to group users to test different recommendation strategies, ensuring that each user uses the same strategy during the test to evaluate the effect.

7. The feed service system according to claim 6, characterized in that: The feed service system adopts a linkage three-level cache mechanism, including: The recommendation data access service module stores the recommendation data through MongoDB, the recommendation real-time correction service module stores the recommendation candidate set in a distributed manner through Elasticsearch; and the recommendation cache service module stores the recommendation candidate set determined to be high frequency through Redis.

8. The feed service system according to claim 7, characterized in that: The recommendation cache service module periodically queries active users within a preset time and stores the recommendation candidate sets of the active users in redis; The redis uses the sorted set zset for storage and uses the user's unique identifier as the cache access key; The recommendation interface service module obtains a recommendation candidate set through the user unique identifier; The media that the user has read and added to the blacklist are cached in redis through the recommendation cache service module, and the set collection is used to cache according to the user's unique identifier, and the media that has been read and added to the blacklist are filtered before returning the recommendation candidate set to the user.

9. An electronic device, comprising a memory and a processor, characterized in that: The memory stores a computer program that can be run on the processor, and when the program is executed on the processor, the steps in the method for constructing a media personalized recommendation feed service system as described in any one of claims 1 to 4 are implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps in the method for constructing a media personalized recommendation feed service system as described in any one of claims 1 to 4 are implemented.

Images