WO2024020708A1 - 基于用户画像的数据处理方法、装置、设备、介质及程序 - Google Patents

基于用户画像的数据处理方法、装置、设备、介质及程序 Download PDF

Info

Publication number
WO2024020708A1
WO2024020708A1 PCT/CN2022/107564 CN2022107564W WO2024020708A1 WO 2024020708 A1 WO2024020708 A1 WO 2024020708A1 CN 2022107564 W CN2022107564 W CN 2022107564W WO 2024020708 A1 WO2024020708 A1 WO 2024020708A1
Authority
WO
WIPO (PCT)
Prior art keywords
user
data
user data
query request
target database
Prior art date
Application number
PCT/CN2022/107564
Other languages
English (en)
French (fr)
Inventor
陈睿
王淑君
杨韬
葛宏莉
彭小勇
Original Assignee
深圳市富途网络科技有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 深圳市富途网络科技有限公司 filed Critical 深圳市富途网络科技有限公司
Priority to PCT/CN2022/107564 priority Critical patent/WO2024020708A1/zh
Priority to CN202280002410.2A priority patent/CN115485676A/zh
Publication of WO2024020708A1 publication Critical patent/WO2024020708A1/zh

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
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2455—Query execution
    • G06F16/24568—Data stream processing; Continuous queries
    • 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
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2462—Approximate or statistical queries
    • 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
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2477—Temporal data queries

Definitions

  • the embodiments of the present application relate to the field of data processing technology, and in particular to a data processing method, device, equipment, medium and program based on user portraits.
  • User portrait is a tool to describe users and connect user demands with product design direction. It is used in fields such as product design and precision marketing.
  • the server can determine the user's behavioral preferences and other user characteristics based on the user's gender, age, page visits, product transactions and other user data, and then generate a user portrait, so that it can explore user needs based on the user portrait, that is, one or more user characteristics, and provide Provide users with more efficient and targeted services.
  • the Kappa architecture is a data processing method that can not only process data in real time, but also implement data replay capabilities based on the data retention function of its message queue, thereby completing offline analysis or recalculation of the data. For example, when the server recalculates user characteristics to generate user portraits, the server can recalculate multiple user data stored in the message queue based on the data replay capability of the Kappa architecture. During the recalculation process, the server can read multiple user data in sequence.
  • the server can calculate the first user data, generate the user characteristics corresponding to the first user data, and store the user characteristics in In the data table, when the second user data is read, the second user data can be calculated to generate the user characteristics corresponding to the second user data, and then the server can use the user characteristics corresponding to the second user data to update The user characteristics corresponding to the first user data in the data table, and so on, the server can complete the recalculation of multiple user data to obtain the user characteristics, thereby generating a user portrait based on the user characteristics.
  • the data storage method in the above calculation process will lead to inaccurate user feature query, which will lead to inaccurate user portraits generated based on the queried user features.
  • the user features are queried during the re-calculation process, obviously at this time
  • the server has not yet completed the recalculation of the above multiple user data, so the user characteristics queried will not be the user characteristics calculated based on the above multiple user data, which will lead to inaccurate query results, thus affecting the generation and application of user portraits. All have a certain impact.
  • This application provides a data processing method, device, equipment, medium and program based on user portraits to solve the problem of inaccurate user feature query caused by data storage in the prior art, resulting in user portraits generated based on the queried user features. Inaccurate problems can improve the accuracy of user feature query, thereby improving the accuracy of generating user portraits, thus improving the efficiency and accuracy of user portrait applications.
  • this application provides a data processing method based on user portraits.
  • the method includes: obtaining multiple user data and the generation time of the last user data in the multiple user data; if each read from the multiple user data A first user data, then generate the user characteristics corresponding to the first user data; if the generation time of the first user data is consistent with the generation time of the last user data, then store the user characteristics corresponding to the first user data in the target database; If the generation time of the first user data is inconsistent with the generation time of the last user data, the user characteristics corresponding to the first user data are not stored in the target database; a user portrait is generated based on the user characteristics in the target database.
  • this application provides a data processing device based on user portraits, including: a first acquisition module, a generation module, a processing module and a second generation module, wherein the first acquisition module is used to acquire multiple user data and multiple The generation time of the last user data among the user data; the generation module is used to generate the user characteristics corresponding to the first user data if each first user data is read from the plurality of user data; the processing module is used to generate the user characteristics corresponding to the first user data if the first user data is read.
  • the generation time of user data is consistent with the generation time of the last user data, then the user characteristics corresponding to the first user data will be stored in the target database; if the generation time of the first user data is inconsistent with the generation time of the last user data, then the user characteristics corresponding to the first user data will not be stored in the target database.
  • the user characteristics corresponding to the first user data are stored in the target database; the second generation module is used to generate user portraits based on the user characteristics in the target database.
  • an electronic device including: a processor and a memory.
  • the memory is used to store a computer program.
  • the processor is used to call and run the computer program stored in the memory to execute the first aspect or its respective implementations. method in.
  • a fourth aspect provides a computer-readable storage medium for storing a computer program, the computer program causing a computer to execute the method as in the first aspect or its respective implementations.
  • a fifth aspect provides a computer program product, including computer program instructions, which cause a computer to execute the method in the first aspect or its respective implementations.
  • a sixth aspect provides a computer program, which causes a computer to execute the method in the first aspect or its respective implementations.
  • the server can first obtain multiple user data and the generation time of the last user data in the multiple user data. If each first user data is read from the multiple user data, the server can generate the first user data. For the user characteristics corresponding to the user data, if the generation time of the first user data is consistent with the generation time of the last user data, the server can store the user characteristics corresponding to the first user data in the target database; if the generation time of the first user data and If the generation time of the last user data is inconsistent, the server may not store the user characteristics corresponding to the first user data in the target database. Finally, the server may generate a user portrait based on the user characteristics in the target database.
  • the server can determine whether the generation time of the currently read user data is consistent with the generation time of the last user data, and then determine whether the currently read user data is the last user data, so that only the last user data can be
  • the user characteristics corresponding to one user data are stored in the target database, but the user characteristics corresponding to other user data are not stored in the target database.
  • the user characteristics are queried, for example, when the server reads the user data, the user characteristics are processed.
  • the result of the query will not be the user characteristics corresponding to other user data, but only the last original corresponding user characteristics, which is the end user characteristics calculated based on all user data, so that the server can generate the correct user characteristics based on the above end user characteristics.
  • user portraits which can solve the problem of inaccurate user feature queries caused by data storage in the existing technology, resulting in inaccurate user portraits generated based on the queried user features, and can improve the accuracy of user feature queries, thereby improving The accuracy of generating user portraits can improve the efficiency and accuracy of user portrait applications.
  • Figure 1 is an application scenario diagram provided by an embodiment of the present application.
  • Figure 2 is a flow chart of a data processing method based on user portraits provided by an embodiment of the present application
  • Figure 3 is a schematic diagram of data processing based on user portraits provided by an embodiment of the present application.
  • Figure 4 is a schematic diagram of another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 5 is a schematic diagram of yet another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 6 is a schematic diagram of another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 7 is a schematic diagram of another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 8 is a schematic diagram of another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 9 is a schematic diagram of another data processing based on user portraits provided by the embodiment of the present application.
  • Figure 10 is a schematic diagram of a data processing device 1000 based on user portraits provided by an embodiment of the present application.
  • Figure 11 is a schematic block diagram of an electronic device provided by an embodiment of the present application.
  • the server can recalculate multiple user data stored in its message queue based on the data replay capability of the Kappa architecture.
  • the server can sequentially Read each user data in multiple user data.
  • the server can calculate the first user data, generate the user characteristics corresponding to the first user data, and add the user The characteristics are stored in the data table.
  • the second user data can be calculated to generate the user characteristics corresponding to the second user data. Then the server can use the second user data corresponding to The user characteristics update the user characteristics corresponding to the first user data in the data table, and by analogy, the server can complete recalculation of multiple user data.
  • the data storage method in the above calculation process will lead to inaccurate user feature query, which will lead to inaccurate user portraits generated based on the queried user features. For example: if the user features are queried during the re-calculation process, obviously at this time The server has not yet completed the recalculation of multiple user data, so the user characteristics queried will not be the user characteristics calculated based on the above multiple user data, resulting in inaccurate query results, which in turn affects the generation and application of user portraits. cause certain impact.
  • the server can first obtain multiple user data and the generation time of the last user data in the multiple user data. If each first user data is read from the multiple user data, the server can generate the first user data. For the user characteristics corresponding to the user data, if the generation time of the first user data is consistent with the generation time of the last user data, the server can store the user characteristics corresponding to the first user data in the target database; if the generation time of the first user data and If the generation time of the last user data is inconsistent, the server may not store the user characteristics corresponding to the first user data in the target database. Finally, the server may generate a user portrait based on the user characteristics in the target database.
  • the server can determine whether the generation time of the currently read user data is consistent with the generation time of the last user data, and then determine whether the currently read user data is the last user data, so that only the last user data can be
  • the user characteristics corresponding to one user data are stored in the target database, but the user characteristics corresponding to other user data are not stored in the target database.
  • the user characteristics are queried, for example, when the server reads the user data, the user characteristics are processed.
  • the result of the query will not be the user characteristics corresponding to other user data, but only the last original corresponding user characteristics, which is the end user characteristics calculated based on all user data, so that the server can generate the correct user characteristics based on the above end user characteristics.
  • user portraits which can solve the problem of inaccurate user feature queries caused by data storage in the existing technology, resulting in inaccurate user portraits generated based on the queried user features, and can improve the accuracy of user feature queries, thereby improving The accuracy of generating user portraits can improve the efficiency and accuracy of user portrait applications.
  • Figure 1 is an application scenario diagram provided by an embodiment of the present application.
  • the application scenario may include a terminal 110 and a server 120.
  • Terminal 110 and server 120 can communicate with each other.
  • the server 120 can recalculate multiple user data stored in its message queue based on the data replay capability of the Kappa architecture, obtain user characteristics, and store the user characteristics in the database, so that it can be based on the user data in the database.
  • the database may be a database inside the server 120 or a database inside the server 120.
  • the server 120 can be a data center. When it is necessary to reconstruct a user profile, such as when the calculation caliber of user data such as stock account opening changes, the data center can re-acquire the user data in the message queue and determine the new user data. User characteristics, and then regenerate user portraits.
  • the user characteristics query client may be installed on the terminal 110, and the user may query the user characteristics based on natural language by accessing the user characteristics query client.
  • the terminal 110 may not be installed with the user characteristics query client, and the user may browse the user characteristics query client by browsing the user characteristics query client.
  • the server 120 can query the above-mentioned user characteristics based on natural language.
  • the server 120 can convert the above-mentioned natural language into a structured query language (Structured Query Language, SQL) corresponding to the natural language, and query the above-mentioned user characteristics based on the SQL corresponding to the natural language.
  • SQL Structured Query Language
  • the terminal 110 can be a mobile phone, a tablet computer, a desktop, a laptop, a handheld computer, a notebook computer, a vehicle-mounted device, an Ultra-Mobile Personal Computer (UMPC), a netbook, and a cellular computer.
  • Telephones, personal digital assistants (Personal Digital Assistant, PDA), augmented reality (Augmented Reality, AR) and virtual reality (Virtual Reality, VR) equipment are not limited by this application.
  • the server 120 may be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The embodiments of this application do not limit this.
  • Figure 2 is a flow chart of a data processing method based on user portraits provided by an embodiment of the present application. This method can be executed by the server 120 as shown in Figure 1. This application does not limit this, as shown in Figure 2. The method may include the following steps:
  • S230 Determine whether the generation time of the first user data is consistent with the generation time of the last user data. If the generation time of the first user data is consistent with the generation time of the last user data, execute S240. If the generation time of the first user data is consistent, If it is inconsistent with the generation time of the last user data, S250 will be executed;
  • S260 Generate user portraits based on user characteristics in the target database.
  • the above steps are steps performed by the server during data replay.
  • the above steps may be steps performed when the server recalculates multiple user data based on the data replay capability of the Kappa architecture to generate user portraits.
  • the Kappa architecture includes a message queue, a stream processing cluster, and a data table.
  • the message queue has a data retention function and can store user data.
  • the stream processing cluster can read user data in the message queue.
  • Kafka is a messaging system with historical data storage and historical data replay functions
  • Kafka can be selected as the message queue.
  • the stream processing cluster can calculate the user data read from the message queue and obtain the user characteristics corresponding to the user data.
  • Flink is a computing framework that supports stream-batch integrated processing, you can choose Flink as the stream processing cluster.
  • Data tables can be used to store user characteristics corresponding to user data calculated by the stream processing cluster.
  • this application first uses the data replay capability of the server based on the Kappa architecture to recalculate multiple user data, taking the generation of user portraits as an example to briefly introduce the above steps.
  • the server can obtain multiple user data from other databases and store the multiple user data in the message queue.
  • the server needs the data replay capability based on the Kappa architecture to recalculate the multiple user data stored in its message queue.
  • the server can obtain the generation time of the last user data among the multiple user data mentioned above. If the data version or generation time is smaller than the stored version, filtering or deleting the process data can ensure that the data is written correctly.
  • the server can sequentially read multiple user data stored in the message queue based on the stream processing cluster.
  • the stream processing cluster can calculate the first user data and generate the first User characteristics corresponding to the user data, and judging that the generation time of the first user data is inconsistent with the generation time of the last user data, the server will not store the user characteristics corresponding to the first user data in the data table, that is, the above-mentioned target database. , then, the server can read the second user data stored in the message queue based on the stream processing cluster. The stream processing cluster can calculate the second user data, generate the user characteristics corresponding to the second user data, and determine the second user data. Whether the generation time of the first user data is inconsistent with the generation time of the last user data.
  • the server can determine that the second user data is the above-mentioned last user data, so the server can determine the user corresponding to the second user data.
  • feature is the user feature corresponding to the last user data, then the server can store the above user feature in the data table; if it is inconsistent, the server can determine that the second user data is not the above last user data, then the server can not store the second user feature in the data table.
  • the user characteristics corresponding to the two user data are stored in the data table. Similarly, for other user data, the server can perform steps similar to the second user data mentioned above.
  • the query results will only be the user characteristics corresponding to the last user data, that is, the latest user data, and not the user characteristics corresponding to other user data, so the query results are accurate. , which can solve the problem of inaccurate user feature query caused by data storage in the existing technology, resulting in inaccurate user portraits generated based on the queried user features, and can improve the accuracy of user feature query, thereby improving user profile applications. efficiency and accuracy.
  • the server when the server builds a user portrait, it can first determine the user's behavioral preferences and other user characteristics based on user data such as the user's gender, age, page visits, product transactions, etc., and then build the user portrait. , so that user needs can be discovered based on user portraits, that is, one or more user characteristics, and more efficient and targeted services can be provided to users. Among them, user characteristics can be stored in the target database. When user needs are discovered based on user characteristics, The corresponding user characteristics can be queried through the user portrait interface service to apply the queried user characteristics to downstream businesses, such as product design, precision marketing and other fields.
  • the user characteristics that the data center needs to determine are: the number of users who have opened stock accounts, and the user data that needs to be used to determine the user characteristics is: the user's stock account opening status.
  • the data center can analyze the above based on the Kappa architecture. User data is calculated to determine user characteristics. Among them, the message queue of the Kappa architecture is Kafka, the stream processing cluster is Flink, and the target database for storing user characteristics is a data table.
  • the calculation caliber is: the account opening status in the statistical user data is the number of user data that has opened an account.
  • the data center can obtain multiple user data about the user's stock account opening from other databases, such as business source databases that record user data such as stock account opening, and store it in Kafka. Assume that the data There are three user data obtained by the center about the user's stock account opening status.
  • the first user data is that user 1's account opening status for stock 1 at 13:22 on June 30, 2022 is open
  • the second user data is The account opening status of user 2 for stock 1 at 13:23 on June 30, 2022 is under review.
  • the third user data is the account opening status of stock 2 by user 3 at 13:30 on June 30, 2022. If you have not opened an account, then the data center can obtain the generation time of the last user data among the multiple user data, that is, the generation time of the third user data: 13:30 on June 30, 2022. Then, the data center can obtain the generation time based on Flink reads the above three user data from Kafka respectively.
  • Flink can determine that the user characteristics corresponding to the first user data, that is, the number of stock account opening users is 1, and can determine that the number of users who have opened stock accounts is 1.
  • the generation time of one user data is 13:22 on June 30, 2022. This is inconsistent with the generation time of the last user data at 13:30 on June 30, 2022.
  • the data center will not store the first user data.
  • the corresponding user characteristics are stored in the data table. Then, the data center can read the second user data from Kafka based on Flink.
  • Flink can determine that the user characteristic corresponding to the second user data is 2, and can determine the The generation time of the two user data is 13:23 on June 30, 2022, which is inconsistent with the generation time of the last user data at 13:30 on June 30, 2022.
  • the data center will not store the second user data.
  • the corresponding user characteristics are stored in the data table.
  • the data center can read the third user data from Kafka based on Flink.
  • Flink can determine that the user characteristic corresponding to the third user data is 2, and can determine the The generation time of the three user data is 13:30 on June 30, 2022, which is consistent with the generation time of the last user data at 13:30 on June 30, 2022.
  • the data center can match the third user data
  • the user characteristics are stored in the data table, so that the data center can determine the user characteristics: the number of stock account opening users is 2.
  • the returned query result will only be the user characteristics calculated based on the last user data, that is, the number of stock account opening users is 2, and because based on other
  • the user characteristics calculated from user data are not stored in the data table, so the query result will not be the user characteristics calculated based on other user data.
  • the number of stock account opening users is 1. Therefore, the accuracy of user characteristics query can be improved, and then It can improve the accuracy and efficiency of generating user portraits.
  • the data center when building user portraits, can first obtain user data from other databases such as business data sources through Binlog synchronization or Agent reporting.
  • This user data can be used in user stocks.
  • Account opening status, user login status and other user data, and the obtained user data is placed in the data operation layer (Operation Data Store, ODS) layer.
  • ODS Operaation Data Store
  • the ODS layer is the closest to the original data in the data source in the data model, that is, user data.
  • Binlog synchronization is a way for the data center to obtain user data from the above-mentioned other databases and store it in the ODS layer.
  • Binlog synchronization includes Changelog (update Log), the data center can capture the changed data in other databases in real time by updating the log, and synchronize the changed data to the ODS layer.
  • a user data is recorded in other databases: User 2’s account opening status for Stock 1 is During the review, the data center can obtain this user data from other databases through Binlog synchronization and store it in the ODS layer. At this time, the user data stored in the ODS layer is that the account opening status of user 2 for stock 1 is under review.
  • the data center can capture the above change data in real time by updating the log, that is, "account opened", then the data center can This change data is synchronized to the ODS layer, that is, the account opening status of user data User 2 stored in the ODS layer for Stock 1 is changed from "under review” to "account opened”. In this way, data transmission overhead can be reduced and improve transmission efficiency.
  • the data center can calculate the user data of the ODS layer at the computing layer to determine the user characteristics.
  • the calculation method can be a stream calculation method or a batch calculation method. The specific calculation process will be described in the following embodiments. A detailed introduction will not be given here.
  • the data center can store user characteristics in the Data Warehouse Service (DWS) layer, and then generate user portraits.
  • DWS Data Warehouse Service
  • the DWS layer can be used to store calculation results, such as calculations for user data in the ODS layer. result. It can be understood that when the above data center calculates the above user data based on the Kappa architecture, Kafka can be used as the ODS layer and the data table can be used as the DWS layer.
  • the server when the server reads a first user data from multiple user data, when generating the user characteristics corresponding to the first user data, it can first select the first user data corresponding to the calculation caliber and application scenario.
  • the user characteristic generation method is a stream generation method or a batch generation method. Then, the server can generate the user characteristics corresponding to the first user data based on the above user characteristic generation method corresponding to the first user data.
  • the data center can complete the flow generation method, that is, the flow calculation method, based on the Kappa architecture. It can trigger the batch generation method, that is, the batch calculation method, through Airflow. Of course, it can also trigger the batch generation method regularly through other methods. This application does not limit this. .
  • Airflow is a task scheduling tool that can set the trigger time of computing tasks such as batch computing tasks and the execution time of computing tasks such as the execution time of batch computing tasks.
  • the server can complete the flow generation method based on the Kappa architecture, and can also complete the flow generation method based on the Lambda architecture, which is not limited in this application.
  • the Lambda architecture is a data processing architecture, including two modules: real-time processing, that is, stream computing, and offline processing, that is, batch computing, so the maintenance cost is high, while the Kappa architecture does not have an offline processing module, that is, a batch computing module, so it is based on the Kappa architecture Completing the flow generation method reduces maintenance costs.
  • the batch generation method that is, the batch calculation method
  • the batch calculation method must first define the calculation job logic and submit it to the churn calculation system, and the calculation job logic cannot be changed during the entire running period.
  • the data calculated by the batch calculation method must be loaded into the calculation system in advance, and the subsequent calculation system will perform calculations after the data loading is completed.
  • the stream generation method that is, the stream computing method, places more emphasis on computing data flow and low latency.
  • the stream computing method can spread a large amount of data to each point in time, continuously transmit small batches, the data continues to flow, and the data is discarded after the calculation is completed.
  • the results calculated by the stream computing method can be immediately delivered to the online system to achieve real-time display.
  • the user feature 1 that needs to be determined is: the number of views of page 1 by user 1 in the 10 days before 13:30 on May 30, 2022; the application scenario 1 of user feature 1 is: predicting users based on user feature 1 1’s number of views on page 1 within 10 days after 13:30 on May 30, 2022, to determine whether the actual number of views on page 1 by user 1 within 10 days after 13:30 on May 30, 2022 is consistent with the above prediction The number of views is consistent; the calculation caliber 1 to determine user characteristics 1 is: obtain user 1’s browsing data of all pages in the 10 days before 13:30 on May 30, 2022, and then count the number of views of page 1 in the data, According to the application scenario 1 and calculation caliber 1 corresponding to the above user feature 1, it can be determined that the real-time requirements for determining user feature 1 are not high, so the data center can choose the batch generation method, that is, the batch calculation method, to generate user feature 1.
  • the data center can use Airflow settings to generate the start time of user feature 1, such as 13:30 on June 30, 2022, and then at that time, obtain user 1's information in the 10 days before 13:30 on May 30, 2022. View data of all pages, and then use the batch calculation method to count the number of views of page 1 in the data, thereby determining user characteristics 1.
  • the user feature 2 that needs to be determined is: whether user 2 has logged into application 1 in the past three days; the application scenario 2 of user feature 2 is: judging whether user 2 is an active user based on user feature 2. If it is determined that user 2 If the user is an active user, then message 1 will be pushed to user 2 in real time. If it is determined that user 2 is not an active user, message 1 will not be pushed to user 2; the calculation caliber 2 of user feature 2 is: obtain the data of user 2's latest login to application 1, Determine whether the generation time of the data is within the past three days. If it is within the past three days, determine the user feature 2 as "1". If not within the past three days, determine the user feature 2 as "0".
  • the application scenario 2 and calculation caliber 2 can determine that the real-time requirements for determining user characteristics 2 are relatively high, so the data center can choose the flow generation method, that is, the flow calculation method, to generate user characteristics 2. For example, assuming that the data center needs to determine whether to push message 1 to user 2 at 13:30 on June 30, 2022, the data center can obtain the data of user 2's latest login to application 1 in real time: User 2 in June 2022 Log in to application 1 at 12:30 on March 30. It can be judged that the data was generated within three days. Then the user characteristic can be determined to be "1", that is, user 2 can be judged to be an active user, so that message 1 can be pushed to user 2 in real time.
  • the target database in this embodiment may include but is not limited to a Remote Dictionary Server (Redis) database, an ElasticSearch database, a cloud database, etc.
  • Redis Remote Dictionary Server
  • ElasticSearch ElasticSearch database
  • cloud database etc.
  • the above data center stores user characteristics in the DWS layer (target database), it can store the user characteristics in databases such as Redis database, ElasticSearch database or cloud server.
  • the Redis database uses a key-value storage method, that is, each record only contains a Key for querying data and a corresponding value for storing data, so it has high real-time requirements and a large amount of query data.
  • Smaller query interfaces such as On-Line Transaction Processing (OLTP) generally choose the Redis database as the query engine; the ElasticSearch database can implement high-performance complex aggregation queries, so it does not require high real-time performance and requires a large amount of query data.
  • Large query interfaces such as OLAP (On-Line Analytical Processing) generally choose the ElasticSearch database as the query engine.
  • Cloud storage is suitable for data with large amounts of data, wide data coverage, and high real-time requirements.
  • the data parameters of the data packet are first obtained.
  • the data parameters may include the query frequency Dat_fre corresponding to the data name, the data volume Dat_voe corresponding to the data packet, and the data corresponding to the data name.
  • the priority Dat_pro; then the attribute parameter Dat_pre corresponding to the data packet is determined based on the data parameters:
  • ⁇ , ⁇ , and ⁇ represent attribute factors obtained based on historical data training
  • Dat_mon represents the preset frequency threshold.
  • the frequency threshold is used to measure the query frequency to determine the corresponding attribute parameter determination method in a targeted manner.
  • the query frequency, priority and data volume of the data are considered in the calculation of attribute parameters, so as to measure the storage method of the data packet based on the above parameters.
  • the storage location corresponding to the data packet is determined based on the attribute parameters.
  • the parameter threshold corresponding to each storage method can be preset, and the storage location corresponding to the data packet is determined based on the parameter threshold corresponding to each storage method.
  • the data center After storing user characteristics in storage media such as Redis data and ElasticSearch databases and cloud servers, the data center can determine the query interface corresponding to the query request based on the query request, so that different query interfaces can query based on the appropriate database. At this time, it can be ensured that the databases corresponding to different query interfaces store user characteristics to improve query efficiency.
  • the specific implementation manner in which the above-mentioned data center can select and determine the query interface corresponding to the query request based on the query request will be described in detail in the following embodiments, and will not be described in detail here.
  • the terminal can send a first user feature query request to the server.
  • the server can respond to the first user feature query request and add the first user feature query request to the target database. Find if the target user profile is stored. If the target user characteristics are stored in the target database, the server can send the target user characteristics to the terminal; if the target user characteristics are not stored in the target database, and the target user characteristics consist of multiple sub-user characteristics in the target database, the server can send the target user characteristics to the terminal.
  • the target user characteristics are decomposed into the above-mentioned multiple sub-user characteristics, and data query is performed based on the above-mentioned multiple sub-user characteristics.
  • the server can determine the target user characteristics based on the queried sub-user characteristics, and send the target user characteristics to the terminal, that is to say , the server can also implement combined judgment using multiple user characteristics.
  • the server does not need to generate target user characteristics.
  • the target database only needs to store sub-user characteristics. That is, the target database can only store atomic characteristics.
  • the server can determine the target based on the sub-user characteristics.
  • User characteristics can reduce the computing cost of the server and the storage cost of the target database, that is, it can reduce the redundancy of user portrait data and repeated development costs.
  • the server can realize the combined judgment of user characteristics based on the query of multiple sub-user characteristics. That is, the server can implement combined judgment using multiple user characteristics, which improves the processing capability of the server and improves the user experience.
  • the target user characteristics that the terminal needs to query are: users who have logged in to application 1 in the past three days and whose account opening status for stock 1 is opened.
  • the target database does not store the above target user characteristics, and the target database stores sub-user characteristics. 1 and sub-user feature 2, respectively: users who have logged in to application 1 in the past three days, and users whose account opening status for stock 1 is opened.
  • the server can Search the above target user characteristics in the database and determine that the target user characteristics are not stored in the target database.
  • the server can determine that the target user characteristics can be composed of the intersection of sub-user characteristics 1 and sub-user characteristics 2. Then the server can separately search for the sub-user characteristics in the target database.
  • User feature 1 and sub-user feature 2 determine that sub-user feature 1 is "user 1, user 2, user 3" and sub-user feature 2 is "user 1, user 2", so the server can determine the target user feature to be "user 1, user 2". User 2", then the server can send the determined target user characteristics to the terminal.
  • the server can query the multiple sub-user characteristics in the target database based on the AST tree and based on the multiple sub-user characteristics.
  • the sub-user characteristics determine the target user characteristics, and the above-mentioned AST tree is transformed from the code regarding the above-mentioned plurality of sub-user characteristics and the structural relationship when the plurality of sub-user characteristics constitute the target user characteristics.
  • the server after receiving the first user feature query request sent by the terminal, the server can respond to the first user feature query request and convert the first user feature query request into a second user feature query with the same meaning. request, and then, in response to the second user characteristic query request, the server may search whether the target user characteristic is stored in the target database. In this way, when the server does not find the user characteristics corresponding to the first user characteristics query request in the target database, it can search the corresponding user in the target database based on the second user characteristics query request with the same meaning as the first user characteristics query request.
  • the server can send the user features corresponding to the second user feature query request to the terminal, thereby improving query efficiency and improving the generation of user portraits. efficiency.
  • the terminal sends a request to the server
  • the first user feature query request sent is used to search for user feature 1 in the target database.
  • the server can respond to the first user feature query request by searching the first user feature query request.
  • the request is converted into a second user feature query request with the same meaning.
  • the second user feature query request is used to search for user feature 2 in the target database.
  • the server can respond to the second user feature query request and search for the user feature in the target database.
  • the server may pre-store the corresponding relationship between the second user feature query request and the first user feature query request that has the same meaning as the second user feature query request, and when converting the first user feature query request into the second user feature query request , can be converted according to the above-mentioned stored correspondence, and this application does not limit this.
  • the server can respond to the first user feature query request by querying the target database for user feature 1 corresponding to the first user feature query request. If user feature 1 is not found, , and then convert the first user feature query request into a second user feature query request, and then in response to the second user feature query request, search whether user feature 2 is stored in the target database. This application does not limit this.
  • the server before responding to the above first user feature query request and searching whether the target user feature is stored in the target database, the server can also perform permission verification on the sender of the first user feature query request.
  • the sender can respond to the first user characteristic query request and search whether the target user characteristics are stored in the target database, so as to improve the security of data query and the security of generating user portraits.
  • the server can obtain the sender's identity, and then the server can determine the sender's identity based on the sender's identity. Permission scope. If the sender's permission range includes the permission to query the target user's characteristics, the server can determine that the sender has passed the permission verification; if the sender's permission scope does not include the permission to query the target user's characteristics, the server can determine that the sender's permission range does not include the permission to query the target user's characteristics. The sender failed the permission verification.
  • the server may pre-store the corresponding relationship between the sender's identity and the sender's authority scope.
  • the sender's authority scope includes user characteristics that the sender can search in the target database. For example: Assume that the first user feature query request is used to query user feature 1 in the target database.
  • the server pre-stores the corresponding relationship between the identification of business party 1, business party 1, and the authority scope 1 of business party 1. Assume that authority range 1 includes: User characteristics 1 and user characteristics 2.
  • the server After receiving the first user characteristics query request sent by the terminal, the server can first determine that the identity of the sender of the first user characteristics query request is sender 1, and then the server can add the above pre-stored If it is found in the corresponding relationship that the authority range of sender 1 includes user characteristic 1, then the server can determine that the authority verification of sender 1 has passed.
  • the server before responding to the above first user feature query request and searching whether the target user feature is stored in the target database, can also determine the query interface corresponding to the first user feature query request. Then, the server can Based on the query interface corresponding to the first user feature query request, in response to the first user feature query request, search whether the target user feature is stored in the target database. For example: when determining the query interface corresponding to the first user feature query request, the server can first determine whether the first user feature query request includes a user identifier. If the first user feature query request includes a user identifier, the server can determine the first user feature query request.
  • the query interface corresponding to the request is an OLTP interface; if the first user characteristics query request does not include the user identifier, it can be determined that the query interface corresponding to the first user characteristics query request is an OLAP interface. It can be understood that when the first user characteristics query request includes a user identification, it can generally be determined that the first user characteristics query request is used to query the user characteristics of the user corresponding to the user identification, and then the data included in the query result can be determined. The amount is small. Based on the above description of the OLTP interface and OLAP interface, the query interface can be selected as the OLTP interface. In this embodiment, the OLTP interface can support second-level query of user characteristic values, aggregate counts of characteristic values, and second-level metadata. Query and return.
  • the server can select an appropriate query interface based on user characteristics query requests to improve data query efficiency and data query reliability, thereby improving the reliability and efficiency of generating user portraits.
  • the server receives the first user feature After the query request is made, it can be determined that the first user feature query request includes the user identification user 1, and then the server can select the query interface corresponding to the first user feature query request as the OLTP interface.
  • the server can determine whether the user meets a certain condition through the OLTP interface, and then deliver corresponding advertising pop-ups to users who meet the condition. For example, assuming that the first user characteristics query request is used to query user characteristics 2 in the target data: whether user 2 has an account, and the first user characteristics query request includes user 2’s identification user 2, then the server receives the first user characteristics query request Finally, it can be determined that the first user feature query request includes the user identification user 2, then the server can select the query interface corresponding to the first user feature query request as the OLTP interface, and the server can query that user feature 2 means that user 2 has opened an account. When it is found that user characteristic 2 has not opened an account, an error (false) is returned to the terminal. In addition, when the server confirms that user characteristic 2 is that user 2 has not opened an account, it can recommend account opening pop-up advertisements to user 2.
  • the server can recommend account opening pop-up advertisements to user 2.
  • the server can determine users who meet certain conditions through the OLAP interface and push corresponding messages to the users. For example, assume that the first user feature query request is used to query user feature 3: active users in the target data. Active users refer to users who have logged in in the past three days. After receiving the first user feature query request, the server can determine the The first user characteristics query request does not include a user identifier, then the server can select the query interface corresponding to the first user characteristics query request as the OLAP interface, and then the server can determine all active users through the OLAP interface and push messages to them.
  • the user characteristics in this embodiment may also include user blocked user lists, blocked user lists, self-selected information, special attention stock lists, etc.
  • the system can filter recommendations based on the first two characteristics. content, using the latter two features to recommend relevant posts to users.
  • the server can query multiple user characteristics of a user through the OLTP interface, and then combine the multiple user characteristics to analyze the user's interest preferences and recommend information articles of interest to the user, where the user characteristics can include attention information. , position information, etc.
  • the server can determine that both include user identification: user 4, then the server can select the above two first user feature query requests
  • the corresponding query interfaces are all OLTP interfaces.
  • the server can query that user feature 4 is that user 4 follows stock 1, and user feature 5 is that user 4 does not follow stock 2. Then the server can analyze one of user 4's interest preferences as: Like If Stock 1 does not like Stock 2, the server can recommend information articles, announcements, related posts, news, forums and other content related to Stock 1 to User 4.
  • the server can determine whether the user meets a certain condition through the OLTP interface, and then issue corresponding rewards to users who meet the condition. For example, assuming that the first user characteristics query request is used to query user characteristics 6 in the target data: whether user 6 deposits money, and the first user characteristics query request includes user 6's identification user 6, then the server receives the first user characteristics query request After that, it can be determined that the first user feature query request includes the user identification user 6, then the server can select the query interface corresponding to the first user feature query request as the OLTP interface, and the server can query the user feature 6 to indicate that user 6 has deposited money. At the time, a deposit reward will be issued to user 6.
  • the server can determine the user characteristics: the user's assets. Whether the target conditions are met to control the user's stock market quotation browsing permissions. For example, assuming that the first user feature query request is used to query user feature 7 in the target data: whether the assets of user 7 are 10,000, then the server can determine the first user feature query request after receiving the first user feature query request.
  • the request includes the user identification user 7, then the server can select the query interface corresponding to the first user feature query request as the OLTP interface, and the server can query user feature 7 when the assets of user 7 have reached 10,000, and send the request to user 7 Open the authority for the market situation of stock 7.
  • the server can determine whether there are abnormal user characteristics in the target database. If there are abnormal user characteristics in the target database, the server can generate prompt information and push the prompt information to prompt the user that there are abnormal user characteristics in the target database. Abnormal user characteristics; if there are no abnormal user characteristics in the target database, the server does not need to generate prompt information, thereby ensuring the accuracy of user characteristics stored in the target database, improving the accuracy of data query results, and thus improving the efficiency of generating user portraits. Accuracy and efficiency.
  • the server can establish a profile monitoring module to build feature models for different user features or user profiles to determine whether there are abnormal user features in the target database, that is, to perform monitoring and alarming. For example, when the server determines whether there are abnormal user characteristics in the target database, it can obtain the first user characteristics stored in the target data at any time and at least one second user characteristic within a preset time period before the any time. Then, the server can perform statistics on the at least one second user characteristic to obtain the distribution range of the first user characteristic.
  • the server can determine that there are abnormal user characteristics in the target database; if the first user characteristic is not within the above distribution range, the server can determine that there are abnormal user characteristics in the target database; If a user characteristic is within the distribution range, the server can determine that there are no abnormal user characteristics in the target database. For example: the server can obtain the first user characteristic stored in the target database at 24:00 on June 30, 2022: the number of users who logged in to application 1 on June 30, 2022. The first user characteristic is a, and obtain the first user characteristic in June 2022.
  • Second user characteristic 1 and second user characteristic 2 respectively: the number of users logging into application 1 on June 29, 2022, June 2022
  • the number of users who logged in to application 1 on the 28th, the second user feature 1 and the second user feature 2 are b and c respectively.
  • the first user feature a is in the distribution range (d-e, d+e)
  • the above user data can be obtained by the server from other databases, such as business source databases that record user data such as stock account opening status and user login status. Then when the user data stored in the above other databases changes, such as If the storage location is changed, the user data obtained by the server will be inaccurate, which will lead to inaccurate generated user characteristics, which will lead to inaccurate data query results.
  • the target database stores the accuracy of user features, optimizes the calculation and change speed of feature data, and performs real-time stream calculations based on business data more quickly.
  • more than 90% of features support second-level real-time updates to improve the accuracy of data query results, thereby improving the accuracy of data query results. Improve the accuracy and efficiency of generating user portraits.
  • the server can convert the natural language into SQL, and query the target user characteristics in the target database based on the SQL.
  • the server can store the correspondence between natural language and SQL in advance, and then convert the natural language into SQL based on the correspondence. For example, as shown in Table 1, assume that the server pre-stores some correspondence between natural language and SQL as shown in Table 1, where, as shown in the second row in Table 1, the natural language "age” and the SQL "SQL1" correspond.
  • SQL natural language SQL age SQL1 and SQL4 SQL2 social days SQL5 20 SQL3 30 SQL6
  • the terminal can send the natural language 1 to the server.
  • the server After the server receives the natural language 1, it can Table 1 converts natural language 1 into corresponding SQL "SQL1, SQL2, SQL3".
  • the server can include: a permission verification module, a portrait service module, and a target database.
  • the terminal can send a first user characteristic query request to the server to query the target user characteristics.
  • the server After receiving the first user feature query request, the permission verification can be performed on the sender of the first user feature query request based on the above permission verification module.
  • the permission verification on the sender passes, the first user feature query request can be responded to.
  • the query interface corresponding to the first user feature query request is determined based on the above-mentioned portrait service module, and then based on the query interface corresponding to the first user feature query request, in response to the first user feature query request, in the target database Search the target user characteristics. If the target user characteristics are stored in the target database, the server can send the target user characteristics to the terminal.
  • the server can include: a parameter verification module, a SQL parsing module, a routing module, and a target database.
  • the terminal can send a first user feature query request to the server for querying the target user.
  • the first user feature query request can be in natural language.
  • the server can first perform permission verification on the sender of the first user feature query request based on the above parameter verification module.
  • the sender passes the permission verification, it can respond to the first user feature query request and parse the first user feature query request in the form of natural language into a first user feature query request in the SQL form based on the above-mentioned SQL parsing module.
  • the query interface corresponding to the first user feature query request may be determined based on the above routing module, and then the server may, based on the query interface corresponding to the first user feature query request, respond to the first user feature query request by searching in the target database whether The target user characteristics are stored. If the target user characteristics are stored in the target database, the server can send the target user characteristics to the terminal.
  • the server can first obtain multiple user data and the generation time of the last user data among the multiple user data. Each time a first user data is read from the user data, the server can generate the user characteristics corresponding to the first user data. If the generation time of the first user data is consistent with the generation time of the last user data, the server can generate the first user data. The user characteristics corresponding to the user data are stored in the target database; if the generation time of the first user data is inconsistent with the generation time of the last user data, the server may not store the user characteristics corresponding to the first user data in the target database. Finally, the server may Generate user portraits based on user characteristics in the target database.
  • the server can determine whether the generation time of the currently read user data is consistent with the generation time of the last user data, and then determine whether the currently read user data is the last user data, so that only the last user data can be
  • the user characteristics corresponding to one user data are stored in the target database, but the user characteristics corresponding to other user data are not stored in the target database. Then when the user characteristics are queried, for example, when the server reads the user data, the user characteristics are processed.
  • the result of the query will not be the user characteristics corresponding to other user data, but only the last original corresponding user characteristics, which is the end user characteristics calculated based on all user data, so that the server can generate based on the above end user characteristics
  • Correct user portraits can solve the problem of inaccurate user feature queries caused by data storage in the existing technology, resulting in inaccurate user profiles generated based on the queried user features, and can improve the accuracy of user feature queries, thereby improving the accuracy of user feature queries. Improve the efficiency and accuracy of user profiling applications.
  • the server may respond to the first user characteristics query request and search whether the target user characteristics are stored in the target database. If the target user characteristics are not stored in the target database, and the target user characteristics are composed of multiple sub-user characteristics in the target database, the server can decompose the target user characteristics into the above-mentioned multiple sub-user characteristics, and perform data query based on the above-mentioned multiple sub-user characteristics. . In this way, the server does not need to generate target user characteristics, and the target database only needs to store sub-user characteristics.
  • the server can determine the target user characteristics based on the sub-user characteristics, thereby reducing the computing cost of the server. , reduce the storage cost of the target database.
  • the server can respond to the first user feature query request by converting the first user feature query request into a second user feature query request with the same meaning, and then, The server may respond to the second user characteristic query request by searching whether the target user characteristic is stored in the target database. In this way, when the server does not find the user characteristics corresponding to the first user characteristics query request in the target database, it can search the corresponding user in the target database based on the second user characteristics query request with the same meaning as the first user characteristics query request.
  • the server can send the user features corresponding to the second user feature query request to the terminal, thereby improving query efficiency and improving the efficiency of generating user portraits. .
  • the server before responding to the first user feature query request and searching whether the target user feature is stored in the target database, the server can also perform permission verification on the sender of the first user feature query request.
  • the permission verification is passed, in response to the first user characteristic query request, it is possible to search whether the target user characteristics are stored in the target database, so as to improve the security of data query and the security of generating user portraits.
  • the server can also determine the query interface corresponding to the first user feature query request. Then, the server can determine the query interface corresponding to the first user feature query request based on the first user feature query request.
  • the query interface corresponding to the user characteristics query request searches whether the target user characteristics are stored in the target database in response to the first user characteristics query request. In this way, the server can select an appropriate query interface based on user characteristics query requests to improve data query efficiency and data query reliability, and improve the efficiency and reliability of generating user portraits.
  • the server can determine whether there are abnormal user characteristics in the target database. If there are abnormal user characteristics in the target database, the server can generate prompt information and push the prompt information to prompt the user that there are abnormal users in the target database. Characteristics; if there are no abnormal user characteristics in the target database, the server does not need to generate prompt information, thereby ensuring the accuracy of user characteristics stored in the target database, improving the accuracy of data query results, and improving the accuracy of generating user portraits.
  • Figure 10 is a schematic diagram of a data processing device 1000 based on user portraits provided by an embodiment of the present application. As shown in Figure 10, the device 1000 includes:
  • the first acquisition module 1001 is used to acquire multiple user data and the generation time of the last user data in the multiple user data;
  • the first generation module 1002 is configured to generate user characteristics corresponding to the first user data each time a first user data is read from a plurality of user data;
  • the processing module 1003 is configured to: if the generation time of the first user data is consistent with the generation time of the last user data, store the user characteristics corresponding to the first user data in the target database; if the generation time of the first user data is the same as the generation time of the last user data, If the generation time of the user data is inconsistent, the user characteristics corresponding to the first user data will not be stored in the target database;
  • the second generation module 1004 is used to generate user portraits based on user characteristics in the target database.
  • the device 1000 also includes: a second acquisition module 1005, a search module 1006, a decomposition module 1007, and a query module 1008, where the second acquisition module 1005 is used to obtain the first user feature query request; the search module 1006 In response to the first user feature query request, search whether the target user feature is stored in the target database; the decomposition module 1007 is used to if the target user feature is not stored in the target database, and the target user feature is composed of multiple sub-users in the target database feature composition, the target user characteristics are decomposed into multiple sub-user characteristics; the query module 1008 is used to perform data query based on the multiple sub-user characteristics.
  • the search module 1006 is specifically configured to respond to the first user feature query request and convert the first user feature query request into a second user feature query request with the same meaning; in response to the second user feature query request, Check whether the target user characteristics are stored in the target database.
  • the device 1000 also includes: a verification module 1009, where the verification module 1009 is used to perform authority verification on the sender of the first user feature query request; the search module 1006 is specifically used to verify the sender's authority.
  • the permission verification is passed, in response to the first user characteristic query request, it is searched whether the target user characteristics are stored in the target database.
  • the verification module 1009 is specifically used to obtain the identity of the sender; determine the scope of authority of the sender based on the identity of the sender; if the scope of authority of the sender includes the authority to query the characteristics of the target user, determine the authority of the sender.
  • the permission verification of the sender passes; if the sender's permission scope does not include the permission to query the characteristics of the target user, it is determined that the permission check of the sender fails.
  • the device 1000 further includes: a determination module 1010, wherein the determination module 1010 is used to determine the query interface corresponding to the first user feature query request; the search module 1006 is specifically configured to determine the query interface corresponding to the first user feature query request based on the first user feature query request.
  • the query interface in response to the first user feature query request, searches whether the target user feature is stored in the target database.
  • the determination module 1010 is specifically configured to determine that the query interface corresponding to the first user feature query request is an OLTP interface if the first user feature query request includes a user identifier; if the first user feature query request does not include When the user is identified, it is determined that the query interface corresponding to the first user characteristic query request is an OLAP interface.
  • the device 1000 also includes: a judgment module 1011, a third generation module 1012, and a push module 1013.
  • the judgment module 1011 is used to judge whether there are abnormal user characteristics in the target database; the third generation module 1012 uses If there are abnormal user characteristics in the target database, prompt information is generated; the push module 1013 is used to push prompt information to remind the user that there are abnormal user characteristics in the target database.
  • the judgment module 1011 is specifically configured to obtain the first user characteristics stored in the target database at any time and at least one second user characteristic within a preset time period before any time; for at least one second user Statistics are performed on the features to obtain the distribution range of the first user feature; if the first user feature is not within the distribution range, it is determined that there are abnormal user features in the target database; if the first user feature is within the distribution range, it is determined that there are no abnormal user features in the target database. There are abnormal user characteristics.
  • the first generation module 1003 is specifically configured to select the generation of user characteristics corresponding to the first user data according to the calculation caliber and application scenario when each first user data is read from multiple user data.
  • the generation method is a stream generation method or a batch generation method; the user characteristics corresponding to the first user data are generated based on the generation method of the user characteristics corresponding to the first user data.
  • the device embodiments and the method embodiments may correspond to each other, and similar descriptions may refer to the method embodiments. To avoid repetition, they will not be repeated here.
  • the device 1000 shown in Figure 10 can execute the above method embodiments, and the foregoing and other operations and/or functions of each module in the device 1000 are respectively to implement the corresponding processes in each of the above methods. For the sake of brevity, they are not mentioned here. Again.
  • the device 1000 in the embodiment of the present application is described above from the perspective of functional modules in conjunction with the accompanying drawings. It should be understood that this functional module can be implemented in the form of hardware, can also be implemented through instructions in the form of software, or can also be implemented through a combination of hardware and software modules. Specifically, each step of the method embodiments in the embodiments of the present application can be completed by integrated logic circuits of hardware in the processor and/or instructions in the form of software. The steps of the methods disclosed in conjunction with the embodiments of the present application can be directly embodied in hardware. The execution of the decoding processor is completed, or the execution is completed using a combination of hardware and software modules in the decoding processor.
  • the software module may be located in a mature storage medium in the field such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, register, etc.
  • the storage medium is located in the memory, and the processor reads the information in the memory and completes the steps in the above method embodiment in combination with its hardware.
  • Figure 11 is a schematic block diagram of an electronic device provided by an embodiment of the present application.
  • the electronic device may include:
  • Memory 1110 and processor 1120 are used to store computer programs and transmit the program code to the processor 1120.
  • the processor 1120 can call and run the computer program from the memory 1110 to implement the method in the embodiment of the present application.
  • the processor 1120 may be configured to execute the above method embodiments according to instructions in the computer program.
  • the processor 1120 may include but is not limited to:
  • DSP Digital Signal Processor
  • ASIC Application Specific Integrated Circuit
  • FPGA Field Programmable Gate Array
  • the memory 1110 includes, but is not limited to:
  • Non-volatile memory can be read-only memory (Read-Only Memory, ROM), programmable read-only memory (Programmable ROM, PROM), erasable programmable read-only memory (Erasable PROM, EPROM), electrically removable memory. Erase programmable read-only memory (Electrically EPROM, EEPROM) or flash memory. Volatile memory may be Random Access Memory (RAM), which is used as an external cache.
  • RAM Random Access Memory
  • RAM static random access memory
  • DRAM dynamic random access memory
  • DRAM synchronous dynamic random access memory
  • SDRAM double data rate synchronous dynamic random access memory
  • Double Data Rate SDRAM DDR SDRAM
  • ESDRAM enhanced synchronous dynamic random access memory
  • SLDRAM synchronous link dynamic random access memory
  • Direct Rambus RAM Direct Rambus RAM
  • the computer program can be divided into one or more modules, and the one or more modules are stored in the memory 1110 and executed by the processor 1120 to complete the tasks provided by this application.
  • the one or more modules may be a series of computer program instruction segments capable of completing specific functions. The instruction segments are used to describe the execution process of the computer program in the electronic device.
  • the electronic device may also include:
  • Transceiver 1130 which may be connected to the processor 1120 or the memory 1110.
  • the processor 1120 can control the transceiver 1130 to communicate with other devices. Specifically, it can send information or data to other devices, or receive information or data sent by other devices.
  • Transceiver 1130 may include a transmitter and a receiver.
  • the transceiver 1130 may further include an antenna, and the number of antennas may be one or more.
  • bus system where in addition to the data bus, the bus system also includes a power bus, a control bus and a status signal bus.
  • This application also provides a computer storage medium on which a computer program is stored.
  • the computer program When the computer program is executed by a computer, the computer can perform the method of the above method embodiment.
  • embodiments of the present application also provide a computer program product containing instructions, which when executed by a computer causes the computer to perform the method of the above method embodiments.
  • the computer program product includes one or more computer instructions.
  • the computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable device.
  • the computer instructions may be stored in or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted over a wired connection from a website, computer, server, or data center (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (such as infrared, wireless, microwave, etc.) to another website, computer, server or data center.
  • the computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center integrated with one or more available media.
  • the available media may be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as digital video discs (DVD)), or semiconductor media (such as solid state disks (SSD)), etc.
  • the disclosed systems, devices and methods can be implemented in other ways.
  • the device embodiments described above are only illustrative.
  • the division of the modules is only a logical function division. In actual implementation, there may be other division methods.
  • multiple modules or components may be combined or may be Integrated into another system, or some features can be ignored, or not implemented.
  • the coupling or direct coupling or communication connection between each other shown or discussed may be through some interfaces, indirect coupling or communication connection of devices or modules, and may be in electrical, mechanical or other forms.
  • Modules described as separate components may or may not be physically separated, and components shown as modules may or may not be physical modules, that is, they may be located in one place, or they may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. For example, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种基于用户画像的数据处理方法、装置、设备、介质及程序，包括：获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间（S210）；若从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征（S220）；若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则将第一用户数据对应的用户特征存储至目标数据库（S240）；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则不将第一用户数据对应的用户特征存储至目标数据库（S250）；基于目标数据库中的用户特征生成用户画像。可以提高用户特征查询的准确性，进而提高生成用户画像的准确性，从而提高用户画像应用的效率和精确性。

Description

基于用户画像的数据处理方法、装置、设备、介质及程序 技术领域
本申请实施例涉及数据处理技术领域，尤其涉及一种基于用户画像的数据处理方法、装置、设备、介质及程序。
背景技术
用户画像是一种描述用户、联系用户诉求与产品设计方向的工具，应用于产品设计、精准营销等领域。服务器可以根据用户性别、年龄、页面访问情况、商品交易情况等用户数据，确定用户的行为喜好等用户特征，进而生成用户画像，从而可以根据用户画像即一个或者多个用户特征发掘用户需求，提供给用户更高效、更有针对性的服务。
Kappa架构是一种数据处理方式，不仅可以对数据进行实时处理，还可以基于其消息队列的数据保留功能，实现数据重放能力，进而完成对数据的离线分析或者再次计算。例如，服务器在重新计算用户特征，以生成用户画像时，服务器可以基于Kappa架构的数据重放能力对消息队列存储的多个用户数据进行再次计算，在再次计算过程中，服务器可以依次读取多个用户数据中的每个用户数据，当读取到第一个用户数据时，服务器可以对第一个用户数据进行计算，生成第一个用户数据对应的用户特征，并将该用户特征存储在数据表中，当读取到第二个用户数据时，可以对第二个用户数据进行计算，生成第二个用户数据对应的用户特征，然后服务器可以使用第二个用户数据对应的用户特征更新数据表中第一个用户数据对应的用户特征，以此类推，服务器可以完成对多个用户数据的再次计算，得到用户特征，从而根据用户特征生成用户画像。
然而，上述计算过程中的数据存储方式会导致用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确，例如：若在再次计算过程中对用户特征进行查询，显然此时服务器还没有完成对上述多个用户数据的再次计算，那么查询到的用户特征不会是根据上述多个用户数据计算出的用户特征，进而导致查询结果不准确，从而对用户画像的生成和应用都造成一定的影响。
发明内容
本申请提供一种基于用户画像的数据处理方法、装置、设备、介质及程序，以解决现有技术中因数据存储导致的用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确的问题，可以提高用户特征查询的准确性，进而提高生成用户画像的准确性，从而提高用户画像应用的效率和精确性。
第一方面，本申请提供一种基于用户画像的数据处理方法，该方法包括：获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间；若从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征；若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则不将第一用户数据对应的用户特征存储至目标数据库；基于目标数据库中的用户特征生成用户画像。
第二方面，本申请提供一种基于用户画像的数据处理装置，包括：第一获取模块、生成模块、处理模块和第二生成模块，其中，第一获取模块用于获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间；生成模块用于若从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征；处理模块用于若第一用户数据的生 成时间和最后一个用户数据的生成时间一致，则将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则不将第一用户数据对应的用户特征存储至目标数据库；第二生成模块用于基于目标数据库中的用户特征生成用户画像。
第三方面，提供一种电子设备，包括：处理器和存储器，该存储器用于存储计算机程序，该处理器用于调用并运行该存储器中存储的计算机程序，执行如第一方面或其各实现方式中的方法。
第四方面，提供一种计算机可读存储介质，用于存储计算机程序，计算机程序使得计算机执行如第一方面或其各实现方式中的方法。
第五方面，提供一种计算机程序产品，包括计算机程序指令，该计算机程序指令使得计算机执行如第一方面或其各实现方式中的方法。
第六方面，提供一种计算机程序，计算机程序使得计算机执行如第一方面或其各实现方式中的方法。
通过本申请技术方案，服务器可以先获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间，若从多个用户数据中每读取一个第一用户数据，则服务器可以生成第一用户数据对应的用户特征，若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则服务器可以将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则服务器可以不将第一用户数据对应的用户特征存储至目标数据库，最后，服务器可以基于目标数据库中的用户特征生成用户画像。在上述过程中，服务器可以通过判断当前读取的用户数据的生成时间是否和最后一个用户数据的生成时间一致，进而可以确定当前读取的用户数据是否是最后一个用户数据，从而可以只将最后一个用户数据对应的用户特征存储在目标数据库中，而不将其他用户数据对应的用户特征存储在目标数据库中，那么在对用户特征进行查询时，例如在服务器读取用户数据时对用户特征进行查询，查询到的结果不会是其他用户数据对应的用户特征，只会是最后一个原始对应的用户特征即是根据所有用户数据计算出来的最终用户特征，从而服务器可以根据上述最终用户特征生成正确的用户画像，从而可以解决现有技术中因数据存储导致的用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确的问题，可以提高用户特征查询的准确性，进而提高生成用户画像的准确性，从而提高用户画像应用的效率和精确性。
附图说明
为了更清楚地说明本发明实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本申请实施例提供的一种应用场景图；
图2为本申请实施例提供的一种基于用户画像的数据处理方法的流程图；
图3为本申请实施例提供的一种基于用户画像的数据处理的示意图；
图4为本申请实施例提供的另一种基于用户画像的数据处理的示意图；
图5为本申请实施例提供的再一种基于用户画像的数据处理的示意图；
图6为本申请实施例提供的又一种基于用户画像的数据处理的示意图；
图7为本申请实施例提供的又一种基于用户画像的数据处理的示意图；
图8为本申请实施例提供的又一种基于用户画像的数据处理的示意图；
图9为本申请实施例提供的又一种基于用户画像的数据处理的示意图；
图10为本申请实施例提供的一种基于用户画像的数据处理装置1000的示意图；
图11是本申请实施例提供的电子设备的示意性框图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都属于本发明保护的范围。
需要说明的是，本发明的说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。应该理解这样使用的数据在适当情况下可以互换，以便这里描述的本发明的实施例能够以除了在这里图示或描述的那些以外的顺序实施。此外，术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或单元的过程、方法、系统、产品或服务器不必限于清楚地列出的那些步骤或单元，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或单元。
如上所述，服务器在重新计算用户特征，以生成用户画像时，服务器可以基于Kappa架构的数据重放能力对其消息队列存储的多个用户数据进行再次计算，在再次计算过程中，服务器可以依次读取多个用户数据中的每个用户数据，当读取到第一个用户数据时，服务器可以对第一个用户数据进行计算，生成第一个用户数据对应的用户特征，并将该用户特征存储在数据表中，当读取到第二个用户数据时，可以对第二个用户数据进行计算，生成第二个用户数据对应的用户特征，然后服务器可以使用第二个用户数据对应的用户特征更新数据表中第一个用户数据对应的用户特征，以此类推，服务器可以完成对多个用户数据的再次计算。然而，上述计算过程中的数据存储方式会导致用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确，例如：若在再次计算过程中对用户特征进行查询，显然此时服务器还没有完成对多个用户数据的再次计算，那么查询到的用户特征不会是根据上述多个用户数据计算出的用户特征，从而导致查询结果不准确，进而对用户画像的生成和应用都造成一定的影响。
为了解决上述技术问题，服务器可以先获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间，若从多个用户数据中每读取一个第一用户数据，则服务器可以生成第一用户数据对应的用户特征，若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则服务器可以将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则服务器可以不将第一用户数据对应的用户特征存储至目标数据库，最后，服务器可以基于目标数据库中的用户特征生成用户画像。在上述过程中，服务器可以通过判断当前读取的用户数据的生成时间是否和最后一个用户数据的生成时间一致，进而可以确定当前读取的用户数据是否是最后一个用户数据，从而可以只将最后一个用户数据对应的用户特征存储在目标数据库中，而不将其他用户数据对应的用户特征存储在目标数据库中，那么在对用户特征进行查询时，例如在服务器读取用户数据时对用户特征进行查询，查询到的结果不会是其他用户数据对应的用户特征，只会是最后一个原始对应的用户特征即是根据所有用户数据计算出来的最终用户特征，从而服务器可以根据上述最终用户特征生成正确的用户画像，从而可以解决现有技术中因数据存储导致的用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确的问题，可以提高用户特征查询的准确性，进而提高生成用户画像的准确性，从而提高用户画像应用的效率和精确性。
应理解的是，本申请技术方案可以应用于如下场景，但不限于：
在一些可实现方式中，图1为本申请实施例提供的一种应用场景图，如图1所示，该应用场景中可以包括终端110、服务器120。终端110与服务器120之间可以通信。
示例性的，服务器120可以基于Kappa架构的数据重放能力对其消息队列存储的多个用户数据进行再次计算，得到用户特征，并将用户特征存储在数据库中，从而可以基于该数据库中的用户特征生成用户画像，该数据库可以是服务器120内部的数据库，也可以是服务器120内部的数据库，本申请对此不做限制。例如，服务器120可以是数据中台，在需要重新构建用户画像时，如在针对股票开户情况等用户数据的计算口径变化时，数据中台可以重新获取消息队列中的用户数据，并确定新的用户特征，进而重新生成用户画像。终端110上可以安装用户特征查询客户端，用户通过访问该用户特征查询客户端，基于自然语言对上述用户特征进行查询，或者，终端110上可以不安装有上述用户特征查询客户端，用户通过浏览器，基于自然语言对上述用户特征进行查询，查询时，服务器120可以将上述自然语言转换为自然语言对应的结构化查询语言(Structured Query Language，SQL)，并基于该自然语言对应的SQL对上述数据库中存储的用户特征进行查询，将查询结果返回给终端110。
在一些可实现方式中，终端110可以为手机、平板电脑、桌面型、膝上型、手持计算机、笔记本电脑、车载设备、超级移动个人计算机(Ultra-Mobile Personal Computer，UMPC)、上网本，以及蜂窝电话、个人数字助理(Personal Digital Assistant，PDA)、增强现实(Augmented Reality，AR)\虚拟现实(Virtual Reality，VR)设备，本申请对此不做限制。服务器120可以是独立的物理服务器，也可以是多个物理服务器构成的服务器集群或者分布式系统，还可以是提供云计算服务的云服务器，本申请实施例对此不做限制。
应该理解，图1中的终端、服务器的数目仅仅是示意性的，实际上，根据实际情况需要可以设置任意数目的终端和服务器，本申请对此不做限制。
在介绍了本申请实施例的应用场景之后，下面将对本申请技术方案进行详细阐述：
图2为本申请实施例提供的一种基于用户画像的数据处理方法的流程图，该方法可以由如图1所示的服务器120执行，本申请对此不做限制，如图2所示，该方法可以包括如下步骤：
S210：获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间；
S220：若从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征；
S230：判断第一用户数据的生成时间和最后一个用户数据的生成时间是否一致，若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则执行S240，若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则执行S250；
S240：将第一用户数据对应的用户特征存储至目标数据库；
S250：不将第一原始数据对应的用户特征存储至目标数据库；
S260：基于目标数据库中的用户特征生成用户画像。
应理解的是，上述步骤是服务器在数据重放时执行的步骤，例如，上述步骤可以是服务器基于Kappa架构的数据重放能力对多个用户数据进行再次计算，以生成用户画像时执行的步骤。需要说明的是，Kappa架构包括消息队列、流处理集群、数据表，其中，消息队列具有数据保留功能，可以存储用户数据，在数据重放时，流处理集群可以读取消息队列中的用户数据。由于卡夫卡(Kafka)是一种具有历史数据保存、历史数据重放功能的消息系统，所以可以选择卡夫卡作为消息队列。流处理集群可以对从消息队列中读取的用户数据进行计算，得到用户数据对应的用户特征。由于Flink是一种支持流批一体处理方式的计算框架，所以可以选择Flink作为流处理集群。数据表可以用来存储流处理集群计算得到的用户数据对应的用户特征。
示例性的，本申请在此先以服务器基于Kappa架构的数据重放能力对多个用户数据进 行再次计算，以生成用户画像时为例，对上述步骤进行简单介绍。首先，服务器可以从其他数据库获取多个用户数据，并将该多个用户数据存储在消息队列中，在服务器需要基于Kappa架构的数据重放能力对其消息队列存储的多个用户数据进行再次计算时，服务器可以获取上述多个用户数据中最后一个用户数据的生成时间，如果数据的版本或者生成时间比存储的版本小，将该过程数据过滤或者删除，可以保证正确地写入的数据。具体的，服务器可以基于流处理集群依次读取消息队列中存储的多个用户数据，当读取到第一个用户数据时，流处理集群可以对第一个用户数据进行计算，生成第一个用户数据对应的用户特征，并判断第一个用户数据的生成时间和最后一个用户数据的生成时间不一致，则服务器不会将第一个用户数据对应的用户特征存储在数据表即上述目标数据库中，接着，服务器可以基于流处理集群读取消息队列中存储的第二个用户数据，流处理集群可以对第二个用户数据进行计算，生成第二个用户数据对应的用户特征，并判断第二个用户数据的生成时间和最后一个用户数据的生成时间是否不一致，若一致，则服务器可以确定该第二个用户数据是上述最后一个用户数据，从而服务器可以确定该第二个用户数据对应的用户特征是最后一个用户数据对应的用户特征，那么服务器可以将上述用户特征存储在数据表中；若不一致，则服务器可以确定该第二个用户数据不是上述最后一个用户数据，那么服务器可以不将第二个用户数据对应的用户特征存储在数据表中，类似的，对于其他用户数据，服务器可以执行和上述第二个用户数据类似的步骤。这样一来，在终端对用户特征进行查询时，查询结果只会是最后一个用户数据即最新的用户数据对应的用户特征，而不会是其他用户数据对应的用户特征，那么查询结果就是准确的，从而可以解决现有技术中因数据存储导致的用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确的问题，可以提高用户特征查询的准确性，进而提高用户画像应用的效率和精确性。
需要说明的是，如图3所示，服务器在构建用户画像时，可以先根据用户性别、年龄、页面访问情况、商品交易情况等用户数据，确定用户的行为喜好等用户特征，进而构建用户画像，从而可以根据用户画像即一个或多个用户特征发掘用户需求，提供给用户更高效、更有针对性的服务，其中，用户特征可以存储在目标数据库中，在根据用户特征发掘用户需求时，可以通过用户画像接口服务查询对应的用户特征，以将查询到的用户特征应用到下游业务中，如产品设计、精准营销等领域。
下面实施例中将以服务器为数据中台为例来对本申请技术方案进行说明。需要说明的是，在下面的实施例中，都是基于数据合规来获取、计算用户数据的，且获取的用户数据都是经过用户授权的，同时也对用户数据、用户特征等进行了加密和保护。
在一些可实现方式中，假设数据中台需要确定的用户特征是：股票开户用户数量，确定该用户特征需要用到的用户数据是：用户的股票开户情况，数据中台可以基于Kappa架构对上述用户数据进行计算，确定用户特征，其中，Kappa架构的消息队列是卡夫卡，流处理集群是Flink，存储用户特征的目标数据库是数据表。在数据中台需要根据用户数据重新计算用户特征时，例如，数据中台之前在确定股票开户用户数量时，计算口径是：统计用户数据中开户状态是已开户的用户数据的数量，但是，数据中台现在确定股票开户用户数量时，计算口径变化为：统计用户数据中开户状态是已开户和审核中的用户数据的数量，那么数据中台需要重新确定股票开户用户数量这一用户特征，具体过程为：首先，数据中台可以从其他数据库如记录用户股票开户情况等用户数据的业务源数据库中获取关于用户股票开户情况的多个用户数据，并将其存储在卡夫卡中，假设数据中台获取到的关于用户股票开户情况的用户数据有三个，第一个用户数据是用户1在2022年6月30日13:22时对股票1的开户状态是已开户，第二个用户数据是用户2在2022年6月30日13:23时对股票1的开户状态是审核中，第三个用户数据是用户3在2022年6月30日13:30时对股票2的开户状态是未开户，接着，数据中台可以获取该多个用户数据中最后一个用户数据的生成时间即第三个用户数据的生成时间：2022年6月30日13:30，然后，数据中台可以 基于Flink从卡夫卡中分别读取上述三个用户数据，当读取到第一个用户数据时，Flink可以确定第一个用户数据对应的用户特征即股票开户用户数量是1，并且可以判断第一个用户数据的生成时间是2022年6月30日13:22，这和最后一个用户数据的生成时间2022年6月30日13:30不一致，则数据中台不会将第一个用户数据对应的用户特征存储在数据表中，接着，数据中台可以基于Flink从卡夫卡中读取第二个用户数据，Flink可以确定第二个用户数据对应的用户特征是2，并且可以判断第二个用户数据的生成时间是2022年6月30日13:23，这和最后一个用户数据的生成时间2022年6月30日13:30不一致，则数据中台不会将第二个用户数据对应的用户特征存储在数据表中，最后，数据中台可以基于Flink从卡夫卡中读取第三个用户数据，Flink可以确定第三个用户数据对应的用户特征是2，并且可以判断第三个用户数据的生成时间是2022年6月30日13:30，这和最后一个用户数据的生成时间2022年6月30日13:30一致，则数据中台可以将第三个用户数据对应的用户特征存储在数据表中，从而数据中台可以确定出用户特征：股票开户用户数量为2。那么对于在上述再次计算用户特征的过程中接收到的查询用户特征的查询请求，返回的查询结果只会是根据最后一个用户数据计算出的用户特征即股票开户用户数量为2，而由于根据其他用户数据计算出的用户特征并没有存储在数据表中，所以查询结果就不会是根据其他用户数据计算出的用户特征如股票开户用户数量为1，因此可以提高用户特征查询的准确性，进而可以提高生成用户画像的准确性和效率。
在一些可实现方式中，如图4所示，数据中台在构建用户画像时，可以先通过Binlog同步或者Agent上报等方式从其他数据库如业务数据源中获取用户数据，该用户数据可以用户股票开户情况、用户登录情况等用户数据，并将获取到的用户数据放在数据运营层(Operation Data Store，ODS)层，其中，ODS层是数据模型中最接近数据源中的原始数据即用户数据的一层，一般可以将数据源中的原始数据原封不动的接入。Binlog同步是数据中台从上述其他数据库中获取用户数据并存储在ODS层的一种方式，Agent上报是上述其他数据库主动发送用户数据给ODS层的一种方式，另外，Binlog同步包括Changelog(更新日志)，数据中台可以通过更新日志的方式实时捕捉其他数据库中的变更数据，并将变更数据同步到ODS层，例如，其他数据库中记录了一个用户数据：用户2对股票1的开户状态为审核中，数据中台可以通过Binlog同步从其他数据库中获取到这个用户数据并存储在ODS层，那么此时ODS层存储的用户数据为用户2对股票1的开户状态为审核中，之后，在其他数据库中记录的上述用户数据变更为：用户2对股票1的开户状态为已开户时，数据中台可以通过更新日志的方式实时捕捉上述变更数据即“已开户”，那么数据中台可以将该变更数据同步到ODS层，即将ODS层中存储的用户数据用户2对股票1的开户状态为审核中的“审核中”变更为“已开户”，如此一来，可以减少数据传输开销，提高传输效率。之后，数据中台可以在计算层对ODS层的用户数据进行计算，确定出用户特征，计算方法可以是流计算方法，也可以是批量计算方法，具体计算过程本申请将在下面的实施例中进行详细介绍，在此先不赘述。接着，数据中台可以将用户特征存储在服务数据层(Data Warehouse Service，DWS)层，进而可以生成用户画像，其中，DWS层可以用来存储计算结果，如针对ODS层中的用户数据的计算结果。可以理解的是，在上述数据中台基于Kappa架构对上述用户数据进行计算中，卡夫卡可以作为ODS层，数据表可以作为DWS层。
在一些可实现方式中，服务器从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征时，可以先按照计算口径和应用场景，选择第一用户数据对应的用户特征的生成方法，生成方法是流生成方法或批量生成方法，然后，服务器可以基于上述第一用户数据对应的用户特征的生成方法生成第一用户数据对应的用户特征。其中，数据中台可以基于Kappa架构完成流生成方法即流计算方法，可以通过Airflow定时触发批量生成方法即批量计算方法，当然也可以通过其他方法定时触发批生成方法，本申请对此不做限制。其中，Airflow是一种任务调度工具，可以设置计算任务如批量计算任务的触 发时间和计算任务的执行时长如批量计算任务的执行时长。需要说明的是，服务器可以基于Kappa架构完成流生成方法，还可以基于Lambda架构完成流生成方法，本申请对此不做限制。其中，Lambda架构是一种的数据处理架构，包括实时处理即流计算和离线处理即批量计算两个模块，所以维护成本较高，而Kappa架构没有离线处理模块即批量计算模块，因此基于Kappa架构完成流生成方法可以降低维护成本。
应理解的是，批量生成方法即批量计算方法是一种批量、高时延、主动发起的计算方法。批量计算方法必须先定义计算作业逻辑，并提交到流失计算系统，且该计算作业逻辑在整个运行期间是不可更改的。批量计算方法计算的数据一定需要预先加载到计算系统，后续计算系统才会在数据加载完成后进行计算。不同于批量计算方法，流生成方法即流计算方法更加强调计算数据流和低时延。流计算方法可以将大量数据平摊到每个时间点上，连续地进行小批量的进行传输，数据持续流动，计算完之后就丢弃该数据。流计算方法计算后的结果可以立刻投递到在线系统，做到实时化展现。
示例性的，假设需要确定的用户特征1是：用户1在2022年5月30日13:30前10天内对页面1的浏览次数；用户特征1的应用场景1是：根据用户特征1预测用户1在2022年5月30日13:30后10天内对页面1的浏览次数，以判断用户1在2022年5月30日13:30后10天内对页面1的真实浏览次数是否和上述预测的浏览次数一致；确定用户特征1的计算口径1是：获取用户1在2022年5月30日13:30前10天内对所有页面的浏览情况数据，然后统计该数据中对页面1的浏览次数，根据上述用户特征1对应的应用场景1和计算口径1可以确定出对确定用户特征1的实时性要求不高，所以数据中台可以选择用户特征1的生成方法是批生成方法即批计算方法，例如，数据中台可以通过Airflow设置生成用户特征1的开始时间如2022年6月30日13:30，然后在该时间时，获取用户1在2022年5月30日13:30前10天内对所有页面的浏览情况数据，然后对该数据采用批计算方法，统计该数据中对页面1的浏览次数，从而确定出用户特征1。
示例性的，假设需要确定的用户特征2是：用户2是否在近三天登录过应用1；用户特征2的应用场景2是：根据用户特征2判断用户2是否为活跃用户，若判断用户2为活跃用户，则向用户2实时推送消息1，若判断用户2不是活跃用户，则不向用户2推送消息1；用户特征2的计算口径2是：获取用户2最近一次登录应用1的数据，判断该数据的生成时间是否在近三天内，若在近三天内，则确定用户特征2为“1”，若不在近三天内，则确定用户特征2为“0”，根据上述用户特征2对应的应用场景2和计算口径2可以确定出对确定用户特征2的实时性要求较高，所以数据中台可以选择用户特征2的生成方法是流生成方法即流计算方法。例如，假设数据中台需要在2022年6月30日13:30判断是否要向用户2推送消息1，则数据中台可以实时获取用户2最近一次登录应用1的数据：用户2在2022年6月30日12:30登录应用1，可以判断该数据的生成时间在三天内，那么可以确定用户特征为“1”，即可以判断用户2为活跃用户，从而可以实时向用户2推送消息1。
在一些可实现方式中，如图4所示，本实施例中的目标数据库可以包括但不限于远程字典服务(Remote Dictionary Server，Redis)数据库、ElasticSearch数据库以及云端数据库等。上述数据中台在将用户特征存储在DWS层(目标数据库)时，可以将用户特征分别存储在Redis数据库、ElasticSearch数据库或者云端服务器等数据库中。
应理解的是，Redis数据库采用的是key-value存储方式，即每一条记录只包含一个用于查询数据的Key以及与之对应的存储数据的value，所以对实时性要求较高、查询数据量较小的查询接口如联机事务处理(On-Line Transaction Processing，OLTP)一般会选择Redis数据库作为查询引擎；ElasticSearch数据库可以实现高性能的复杂聚合查询，所以对实时性要求不高、查询数据量较大的查询接口如联机分析处理OLAP(OLAP，On-Line Analytical Processing一般会选择ElasticSearch数据库作为查询引擎。而云端存储适用于数据量较大、 数据涵盖范围广、实时性也要求较高的数据。具体的，在确定一个数据包的存储方式时，本实施例中先获取数据包的数据参数，数据参数可以包括数据名称对应的查询频率Dat_fre、数据包对应的数据量Dat_voe、数据名称对应的数据优先级Dat_pro；之后基于数据参数确定数据包对应的属性参数Dat_pre为：
Figure PCTCN2022107564-appb-000001
其中，α、γ、∈表示根据历史数据训练得到属性因子，Dat_mon表示预设的频率阈值，通过频率阈值衡量查询频率的高低，以针对性的确定对应的属性参数确定方式。本实施例中将数据的查询频率、优先级以及数据量考虑进行属性参数的计算中，以基于上述参数来衡量数据包的存储方式。之后基于属性参数来确定数据包对应的存储位置，具体可以预设各存储方式对应的参数阈值，根据各存储方式对应的参数阈值确定数据包对应的存储位置。通过上述方式保证对数据存储的个性化存储，进而提高数据存储和数据调用的效率，降低数据存储的成本。
在将用户特征分别存储在Redis数据和ElasticSearch数据库以及云端服务器等存储介质之后，数据中台可以基于查询请求选择确定该查询请求对应的查询接口，从而使得不同的查询接口在基于合适的数据库进行查询时，可以保证不同的查询接口对应的数据库都存储有用户特征，以提高查询效率。其中，上述数据中台可以基于查询请求选择确定该查询请求对应的查询接口的具体实现方式将在下面的实施例中进行详细描述，本申请在此先不赘述。
在一些可实现方式中，如图5所示，终端可以向服务器发送第一用户特征查询请求，服务器接收到第一用户特征查询请求后，可以响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。若目标数据库中存储有目标用户特征，则服务器可以将目标用户特征发送给终端；若目标数据库中未存储目标用户特征，且目标用户特征由目标数据库中的多个子用户特征构成，则服务器可以将目标用户特征分解为上述多个子用户特征，并基于上述多个子用户特征进行数据查询，最后，服务器可以根据查询到的子用户特征确定目标用户特征，并将目标用户特征发送给终端，也就是说，服务器还可以实现使用多个用户特征进行组合判断。如此一来，服务器可以不用生成目标用户特征，目标数据库只需要存储子用户特征，也即目标数据库可以只存储原子特征，在服务器收到第一查询请求时，服务器可以根据子用户特征来确定目标用户特征，从而可以减少服务器的计算成本，减少目标数据库的存储成本，也即可以减少用户画像数据的冗余和重复的开发成本。而且在上述过程中，服务器可以基于对多个子用户特征的查询实现用户特征的组合判断，即服务器可以实现使用多个用户特征进行组合判断，提高了服务器的处理能力，提高了用户体验。
示例性的，假设终端需要查询的目标用户特征是：近三天内有登录应用1且对股票1的开户状态为已开户的用户，目标数据库没有存储上述目标用户特征，目标数据库存储有子用户特征1和子用户特征2，分别是：近三天内有登录应用1的用户、对股票1的开户状态为已开户的用户，在服务器接收到终端发送的第一用户特征查询请求后，服务器可以在目标数据库中查找上述目标用户特征，确定目标数据库中未存储该目标用户特征，而服务器可以确定该目标用户特征可以由子用户特征1和子用户特征2的交集构成，那么服务器可以在目标数据库中分别查找子用户特征1和子用户特征2，确定子用户特征1为“用户1、用户2、用户3”、子用户特征2为“用户1、用户2”，从而服务器可以确定目标用户特征为“用户1、用户2”，然后，服务器可以将确定的目标用户特征发送给终端。其中，服务器在确定目标数据库中未存储目标用户特征，确定目标用户特征可以由目标数据库中存储的多个子用户特征构成后，可以基于AST树在目标数据库中查询该多个子用户特征并根据该多个子用户特征确定目标用户特征，上述AST树是由关于上述多个子用户特征、多个子 用户特征构成目标用户特征时的构成关系的代码转化来的。
在一些可实现方式中，服务器在接收到终端发送的第一用户特征查询请求后，可以响应于该第一用户特征查询请求，将第一用户特征查询请求转换为含义相同的第二用户特征查询请求，然后，服务器可以响应于第二用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。如此一来，在服务器没有在目标数据库中查找到第一用户特征查询请求对应的用户特征时，可以根据和第一用户特征查询请求含义相同的第二用户特征查询请求在目标数据库查找对应的用户特征，在目标数据库中存储有第二用户特征查询请求对应的用户特征时，那么服务器可以将该第二用户特征查询请求对应的用户特征发送给终端，从而可以提高查询效率，提高用户画像的生成效率。
示例性的，假设目标数据库未存储用户特征1：用户1是否在近三天有登录应用1，目标数据库存储有用户特征2：用户1最近一个登录应用1的时间在近三天内，终端向服务器发送的第一用户特征查询请求是用来在目标数据库中查找用户特征1的，服务器在接收到第一用户特征查询请求后，可以响应于该第一用户特征查询请求，将第一用户特征查询请求转换为含义相同的第二用户特征查询请求，该第二用户特征查询请求用以在目标数据库中查找用户特征2，然后，服务器可以响应于第二用户特征查询请求，在目标数据库查找用户特征2，并将查询到的用户特征2发送给终端。其中，服务器可以预先存储第二用户特征查询请求和与第二用户特征查询请求含义相同的第一用户特征查询请求的对应关系，在将第一用户特征查询请求转换为第二用户特征查询请求时，可以根据上述存储的对应关系来进行转换，本申请对此不做限制。当然，服务器在接收到第一用户特征查询请求后，可以响应于该第一用户特征查询请求，在目标数据库中查询第一用户特征查询请求对应的用户特征1，在没有查找到用户特征1后，再将第一用户特征查询请求转换为第二用户特征查询请求，然后响应于第二用户特征查询请求，在目标数据库中查找是否存储有用户特征2，本申请对此不做限制。
在一些可实现方式中，服务器在响应于上述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征之前，还可以对第一用户特征查询请求的发送方进行权限校验，在对该发送方进行权限校验通过时，可以响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征，以提高数据查询的安全性，提高生成用户画像的安全性。
示例性的，在服务器响应于上述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征之前，服务器可以获取发送方的标识，然后，服务器可以根据发送方的标识确定发送方的权限范围，若发送方的权限范围包括查询目标用户特征的权限，则服务器可以确定对发送方进行权限校验通过；若发送方的权限范围不包括查询目标用户特征的权限，则服务器可以确定对发送方进行权限校验未通过。其中，服务器可以预先存储有发送方的标识和发送方的权限范围的对应关系，该发送方的权限范围包括发送方可以在目标数据库中查找的用户特征。例如：假设第一用户特征查询请求用于在目标数据库中查询用户特征1，服务器预先存储有业务方1的标识业务方1和业务方1的权限范围1的对应关系，假设权限范围1包括：用户特征1、用户特征2，服务器在接收到终端发送的第一用户特征查询请求后，可以先确定第一用户特征查询请求的发送方的标识为发送方1，然后服务器可以在上述预先存储的对应关系中查找到发送方1的权限范围包括用户特征1，那么服务器可以确定对发送方1进行权限校验通过。
在一些可实现方式中，服务器在响应于上述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征之前，还可以确定第一用户特征查询请求对应的查询接口，然后，服务器可以基于第一用户特征查询请求对应的查询接口，响应于所述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。例如：服务器在确定第一用户特征查询请求对应的查询接口时，可以先判断第一用户特征查询请求是否包括用户标识， 若第一用户特征查询请求包括用户标识，则可以确定第一用户特征查询请求对应的查询接口为OLTP接口；若第一用户特征查询请求不包括用户标识，则可以确定第一用户特征查询请求对应的查询接口为OLAP接口。可以理解的是，在第一用户特征查询请求包括用户标识时，一般可以确定第一用户特征查询请求是用以查询关于该用户标识对应的用户的用户特征的，那么可以确定查询结果包括的数据量较小，结合上述对OLTP接口和OLAP接口的描述，因此可以选择查询接口为OLTP接口，本实施例中基于OLTP接口可以支持秒级查询用户特征值，特征值的聚合计数，秒级元数据查询及返回。同理，在第一用户特征查询请求不包括用户标识时，一般可以确定第一用户特征查询请求是用以查询较为复杂的用户特征的，如近三天内有登录应用1的所有用户，那么可以选择查询接口为OLAP接口。如此一来，服务器可以根据用户特征查询请求，选择合适的查询接口，以提高数据查询效率和数据查询可靠性，进而提高生成用户画像的可靠性和效率。
示例性的，假设第一用户特征查询请求用以在目标数据中查询用户特征1：用户1的年龄，第一用户特征查询请求包括用户1的标识用户1，那么服务器在接收到第一用户特征查询请求后，可以确定该第一用户特征查询请求包括用户标识用户1，那么服务器可以选择该第一用户特征查询请求对应的查询接口为OLTP接口。通过上述方式可以实现支持通过画像锚定多类型人群数据，实现多样化投放需求，例如设定实时人群标签、查找例行或者静态用户群等等。同时可以支持更灵活的锚定人群方式，网关层能转译更多类型的SQL语句。例如支持动态范围的实时数据查询，业务能更灵活的在前端使用画像底层特征数据。
示例性的，服务器可以通过OLTP接口判断用户是否满足某种条件，然后对满足该条件的用户下发对应的广告弹窗。例如，假设第一用户特征查询请求用以在目标数据中查询用户特征2：用户2是否开户，第一用户特征查询请求包括用户2的标识用户2，那么服务器在接收到第一用户特征查询请求后，可以确定该第一用户特征查询请求包括用户标识用户2，那么服务器可以选择该第一用户特征查询请求对应的查询接口为OLTP接口，并且服务器可以在查询到用户特征2是用户2已开户时，向终端返回正确(true)，在查询到用户特征2是用户2未开户时，向终端返回错误(false)。另外，服务器可以在确认用户特征2是用户2未开户时，向用户2进行开户弹窗广告推荐。
示例性的，在新股开盘的时候，本实施例中通过获取用户授权的自选股信息或者关注信息进行查询匹配，针对自选股列表中含有该自选股的客户，拉取“用户自选股列表＝xxxx”的全量用户人群，进行新股开盘的信息推送。
示例性的，服务器可以通过OLAP接口确定出符合某个条件的用户，对该用户进行相应的消息推送。例如，假设第一用户特征查询请求用以在目标数据中查询用户特征3：活跃用户，活跃用户指近三天登录过的用户，那么服务器在接收到第一用户特征查询请求后，可以确定该第一用户特征查询请求没有包括用户标识，那么服务器可以选择该第一用户特征查询请求对应的查询接口为OLAP接口，然后服务器可以通过OLAP接口确定出所有的活跃用户，并对其进行消息推送。
示例性的，本实施例中的用户特征可以还可以包括用户屏蔽用户列表、拉黑用户列表、自选信息、特别关注股票列表等等，当用户进入牛牛圈，系统可以根据前两个特征过滤推荐的内容，使用后两个特征用于推荐相关的帖子给用户。
示例性的，服务器可以通过OLTP接口查询某个用户的多个用户特征，然后结合该多个用户特征，分析该用户的兴趣偏好，给用户推荐感兴趣的资讯文章，其中用户特征可以包括关注信息、持仓信息等等。例如，假设第一用户特征查询请求1用以在目标数据中查询用户特征4：用户4是否关注或者持仓股票1，第一用户特征查询请求2用以在目标数据中查询用户特征5：用户4是否关注股票2，那么服务器在接收到第一用户特征查询请求1和第一用户特征查询请求2后，可以确定都包括用户标识：用户4，那么服务器可以选择上述两个第一用户特征查询请求对应的查询接口都为OLTP接口，然后，服务器可以查询 到用户特征4是用户4关注股票1，用户特征5是用户4没有关注股票2，那么服务器可以分析出用户4的一个兴趣偏好为：喜欢股票1，不喜欢股票2，则服务器可以向用户4推荐和股票1有关的资讯文章、公告、相关帖子、新闻以及论坛等内容。
示例性的，服务器可以通过OLTP接口判断用户是否满足某种条件，然后对满足该条件的用户下发对应的奖励。例如，假设第一用户特征查询请求用以在目标数据中查询用户特征6：用户6是否入金，第一用户特征查询请求包括用户6的标识用户6，那么服务器在接收到第一用户特征查询请求后，可以确定该第一用户特征查询请求包括用户标识用户6，那么服务器可以选择该第一用户特征查询请求对应的查询接口为OLTP接口，并且服务器可以在查询到用户特征6是用户6已入金时，向用户6下发入金奖励。
示例性的，不同用户对应的股票市场权限不一样，比如高资产用户有更高的股票市场行情浏览权限，未开户的用户只有指定的行情浏览权限，所以服务器可以通过确定用户特征：用户的资产是否达到目标条件，来控制用户的股票市场行情浏览权限。例如，假设第一用户特征查询请求用以在目标数据中查询用户特征7：用户7的资产是否为1万，那么服务器在接收到第一用户特征查询请求后，可以确定该第一用户特征查询请求包括用户标识用户7，那么服务器可以选择该第一用户特征查询请求对应的查询接口为OLTP接口，并且服务器可以在查询到用户特征7是用户7已的资产已达到1万，并向用户7开通关于股票7的行情情况的权限。
在一些可实现方式中，服务器可以判断目标数据库中是否存在异常的用户特征，若目标数据库中存在异常的用户特征，则服务器可以生成提示信息，并推送该提示信息，以提示用户目标数据库中存在异常的用户特征；若目标数据库中未存在异常的用户特征，则服务器可以不生成提示信息，从而可以保证目标数据库存储用户特征的准确性，提高数据查询结果的准确性，进而提高生成用户画像的准确性和效率。
示例性的，服务器可以建立一个画像监控模块，来对不同的用户特征或者用户画像建立特征模型，以判断目标数据库中是否存在异常的用户特征，即进行监控告警。例如，服务器在判断目标数据库中是否存在异常的用户特征时，可以获取目标数据中在任一时刻存储的第一用户特征以及在该任一时刻之前的预设时长内的至少一个第二用户特征，然后，服务器可以对上述至少一个第二用户特征进行统计，得到第一用户特征的分布范围，若第一用户特征不在上述分布范围内，则服务器可以确定目标数据库中存在异常的用户特征；若第一用户特征在所述分布范围内，则服务器可以确定目标数据库中不存在异常的用户特征。例如：服务器可以获取目标数据库中在2022年6月30日24:00存储的第一用户特征：2022年6月30日登录应用1的用户数量，该第一用户特征为a，获取2022年6月30日24:00之前的两天内的两个第二用户特征：第二用户特征1和第二用户特征2，分别为：2022年6月29日登录应用1的用户数量、2022年6月28日登录应用1的用户数量，该第二用户特征1和第二用户特征2分别为b、c，然后，服务器可以计算第二用户特征1和第二用户特征2的平均值为(b+c)/2＝d，接着，服务器可以将(d-e，d+e)确定为第一用户特征的分布范围，若第一用户特征a在分布范围(d-e，d+e)中，则可以确定第一用户特征为正常数据，即可以确定目标数据库中不存在异常的用户特征；若第一用户特征a不在分布范围(d-e，d+e)中，则可以确定第一用户特征为异常数据，即可以确定目标数据库中存在异常的用户特征，其中，a、b、c、d、e为正整数。
可以理解的是，由于用户画像生产流程的链路较长，而业务数据迁徙或者数据源变更很容易导致获取的用户数据不准确，从而导致用户画像不准确，所以需要对用户特征或者用户画像进行监控告警。例如，上述用户数据可以是服务器从其他数据库如记录用户股票开户情况、用户登录情况等用户数据的业务源数据库中获取的，那么在上述其他数据库中存储的用户数据发生变更时，如用户数据的存储位置发生了变更，服务器获取到到用户数据就会不准确，进而导致生成的用户特征不准确，从而导致数据查询结果不准确，因此通 过上述判断目标数据库中是否存在异常的用户特征，可以保证目标数据库存储用户特征的准确性，优化了特征数据的计算变更速度，更快速根据业务数据进行实时流计算，目前90％以上的特征支持秒级实时更新，以提高数据查询结果的准确性，进而提高生成用户画像的准确性和效率。
在一些可实现方式中，如图1所示，假设终端110上安装应用1，用户通过访问应用1，基于自然语言对目标用户特征进行查询，在用户在应用1上输入用以查询目标用户特征的自然语言后，服务器可以将该自然语言转换为SQL，并基于该SQL在目标数据库中查询上述目标用户特征。其中，服务器可以预先存储自然语言和SQL的对应关系，然后基于该对应关系，将自然语言转换为SQL。例如，如表1所示，假设服务器预先存储有自然语言和SQL的部分对应关系如表1所示，其中，如表1中的第二行所示，自然语言“年龄”和SQL“SQL1”对应。
表1
自然语言 SQL 自然语言 SQL
年龄 SQL1 且 SQL4
＝ SQL2 社交天数 SQL5
20 SQL3 30 SQL6
示例性的，如图6所示，假设用户在应用1上输入的自然语言1为“年龄＝20”，那么终端可以将自然语言1发送给服务器，服务器接收到自然语言1后，可以基于上述表1将自然语言1转换为对应的SQL“SQL1，SQL2，SQL3”。
示例性的，如图7所示，自然语言1“年龄＝20”后有一个加号按钮，当用户点击了该加号按钮后，可以继续输入自然语言2“社交天数＝20”，且输入的自然语言2和自然语言1的关系如图4所示的“且”所示，表示用户希望查询“年龄＝20且社交天数＝20的用户”，然后，服务器可以将用户输入在自然语言转换为对应的SQL“SQL1，SQL2，SQL3，SQL4，SQL5，SQL2，SQL3”。
应理解的是，本申请对自然语言和SQL的对应关系仅是示意性的，将该自然语言转换为对应的SQL也是示意性的。
在一些可实现方式中，如图8所示，服务器可以包括：权限校验模块、画像服务模块、目标数据库，终端可以向服务器发送第一用户特征查询请求，以用于查询目标用户特征，服务器接收到第一用户特征查询请求后，可以先基于上述权限校验模块对第一用户特征查询请求的发送方进行权限校验，在对该发送方进行权限校验通过时，可以响应于第一用户特征查询请求，基于上述画像服务模块确定第一用户特征查询请求对应的查询接口，然后基于第一用户特征查询请求对应的查询接口，响应于所述第一用户特征查询请求，在目标数据库中查找目标用户特征，若目标数据库中存储有目标用户特征，则服务器可以将目标用户特征发送给终端。
在一些可实现方式中，如图9所示，服务器可以包括：参数校验模块、SQL解析模块、路由模块、目标数据库，终端可以向服务器发送第一用户特征查询请求，以用于查询目标用户特征，第一用户特征查询请求可以是自然语言，服务器接收到第一用户特征查询请求后，可以先基于上述参数校验模块对第一用户特征查询请求的发送方进行权限校验，在对该发送方进行权限校验通过时，可以响应于第一用户特征查询请求，基于上述SQL解析模块将自然语言形式的第一用户特征查询请求解析为SQL形式的第一用户特征查询请求，接着，服务器可以基于上述路由模块确定第一用户特征查询请求对应的查询接口，然后，服务器可以基于第一用户特征查询请求对应的查询接口，响应于所述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征，若目标数据库中存储有目标用户特征，则服务器可以将目标用户特征发送给终端。
综上所述，上述实施例提供的技术方案至少带来以下有益效果：通过本申请技术方案， 服务器可以先获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间，若从多个用户数据中每读取一个第一用户数据，则服务器可以生成第一用户数据对应的用户特征，若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则服务器可以将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则服务器可以不将第一用户数据对应的用户特征存储至目标数据库，最后，服务器可以基于目标数据库中的用户特征生成用户画像。在上述过程中，服务器可以通过判断当前读取的用户数据的生成时间是否和最后一个用户数据的生成时间一致，进而可以确定当前读取的用户数据是否是最后一个用户数据，从而可以只将最后一个用户数据对应的用户特征存储在目标数据库中，而不将其他用户数据对应的用户特征存储在目标数据库中，那么在对用户特征进行查询时，例如在服务器读取用户数据时对用户特征进行查询，查询到的结果不会是其他用户数据对应的用户特征，只会是最后一个原始对应的用户特征即是根据所有用户数据计算出来的最终用户特征，，从而服务器可以根据上述最终用户特征生成正确的用户画像，从而可以解决现有技术中因数据存储导致的用户特征查询不准确，从而导致根据查询到的用户特征生成的用户画像不准确的问题，可以提高用户特征查询的准确性，进而提高用户画像应用的效率和精确性。
进一步的，服务器接收到第一用户特征查询请求后，可以响应于第一用户特征查询请求，可以在目标数据库中查找是否存储有目标用户特征。若目标数据库中未存储目标用户特征，且目标用户特征由目标数据库中的多个子用户特征构成，则服务器可以将目标用户特征分解为上述多个子用户特征，并基于上述多个子用户特征进行数据查询。如此一来，服务器可以不用生成目标用户特征，目标数据库只需要存储子用户特征，在服务器收到第一查询请求时，服务器可以根据子用户特征来确定目标用户特征，从而可以减少服务器的计算成本，减少目标数据库的存储成本。
更进一步的，服务器在接收到终端发送的第一用户特征查询请求后，可以响应于第一用户特征查询请求，将第一用户特征查询请求转换为含义相同的第二用户特征查询请求，然后，服务器可以响应于第二用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。如此一来，在服务器没有在目标数据库中查找到第一用户特征查询请求对应的用户特征时，可以根据和第一用户特征查询请求含义相同的第二用户特征查询请求在目标数据库查找对应的用户特征，在目标数据库中存储有第二用户特征查询请求对应的用户特征时，服务器可以将该第二用户特征查询请求对应的用户特征发送给终端，从而可以提高查询效率，提高生成用户画像的效率。
再进一步的，服务器在响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征之前，还可以对第一用户特征查询请求的发送方进行权限校验，在对该发送方进行权限校验通过时，可以响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征，以提高数据查询的安全性，提高生成用户画像的安全性。
又进一步的，服务器在响应于上述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征之前，还可以确定第一用户特征查询请求对应的查询接口，然后，服务器可以基于第一用户特征查询请求对应的查询接口，响应于所述第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。如此一来，服务器可以根据用户特征查询请求，选择合适的查询接口，以提高数据查询效率和数据查询可靠性，提高生成用户画像的效率和可靠性。
又进一步的，服务器可以判断目标数据库中是否存在异常的用户特征，若目标数据库中存在异常的用户特征，则服务器可以生成提示信息，并推送该提示信息，以提示用户目标数据库中存在异常的用户特征；若目标数据库中未存在异常的用户特征，则服务器可以不生成提示信息，从而可以保证目标数据库存储用户特征的准确性，以提高数据查询结果的准确性，提高生成用户画像的准确性。
图10为本申请实施例提供的一种基于用户画像的数据处理装置1000的示意图，如图10所示，该装置1000包括：
第一获取模块1001，用于获取多个用户数据以及多个用户数据中最后一个用户数据的生成时间；
第一生成模块1002，用于若从多个用户数据中每读取一个第一用户数据，则生成第一用户数据对应的用户特征；
处理模块1003，用于：若第一用户数据的生成时间和最后一个用户数据的生成时间一致，则将第一用户数据对应的用户特征存储至目标数据库；若第一用户数据生成时间和最后一个用户数据的生成时间不一致，则不将第一用户数据对应的用户特征存储至目标数据库；
第二生成模块1004，用于基于所述目标数据库中的用户特征生成用户画像。
在一些可实现方式中，装置1000还包括：第二获取模块1005、查找模块1006、分解模块1007、查询模块1008，其中，第二获取模块1005用于获取第一用户特征查询请求；查找模块1006用于响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征；分解模块1007用于若目标数据库中未存储目标用户特征，且目标用户特征由目标数据库中的多个子用户特征构成，则将目标用户特征分解为多个子用户特征；查询模块1008用于基于多个子用户特征进行数据查询。
在一些可实现方式中，查找模块1006具体用于响应于第一用户特征查询请求，将第一用户特征查询请求转换为含义相同的第二用户特征查询请求；响应于第二用户特征查询请求，在目标数据库中查找是否存储有所目标用户特征。
在一些可实现方式中，装置1000还包括：校验模块1009，其中，校验模块1009用于对第一用户特征查询请求的发送方进行权限校验；查找模块1006具体用于在对发送方进行权限校验通过时，响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。
在一些可实现方式中，校验模块1009具体用于获取发送方的标识；根据发送方的标识确定发送方的权限范围；若发送方的权限范围包括查询目标用户特征的权限，则确定对发送方进行权限校验通过；若发送方的权限范围不包括查询目标用户特征的权限，则确定对发送方进行权限校验未通过。
在一些可实现方式中，装置1000还包括：确定模块1010，其中，确定模块1010用于确定第一用户特征查询请求对应的查询接口；查找模块1006具体用于基于第一用户特征查询请求对应的查询接口，响应于第一用户特征查询请求，在目标数据库中查找是否存储有目标用户特征。
在一些可实现方式中，确定模块1010具体用于若第一用户特征查询请求包括用户标识时，则确定第一用户特征查询请求对应的查询接口为OLTP接口；若第一用户特征查询请求不包括用户标识时，则确定第一用户特征查询请求对应的查询接口为OLAP接口。
在一些可实现方式中，装置1000还包括：判断模块1011、第三生成模块1012、推送模块1013，其中，判断模块1011用于判断目标数据库中是否存在异常的用户特征；第三生成模块1012用于若目标数据库中存在异常的用户特征，则生成提示信息；推送模块1013用于推送提示信息，以提示用户目标数据库中存在异常的用户特征。
在一些可实现方式中，判断模块1011具体用于获取目标数据库中在任一时刻存储的第一用户特征以及在任一时刻之前的预设时长内的至少一个第二用户特征；对至少一个第二用户特征进行统计，得到第一用户特征的分布范围；若第一用户特征不在分布范围内，则确定目标数据库中存在异常的用户特征；若第一用户特征在分布范围内，则确定目标数据库中不存在异常的用户特征。
在一些可实现方式中，第一生成模块1003具体用于若从多个用户数据中每读取一个第 一用户数据，则按照计算口径和应用场景，选择第一用户数据对应的用户特征的生成方法，生成方法是流生成方法或批量生成方法；基于第一用户数据对应的用户特征的生成方法生成第一用户数据对应的用户特征。
应理解的是，装置实施例与方法实施例可以相互对应，类似的描述可以参照方法实施例。为避免重复，此处不再赘述。具体地，图10所示的装置1000可以执行上述方法实施例，并且装置1000中的各个模块的前述和其它操作和/或功能分别为了实现上述各个方法中的相应流程，为了简洁，在此不再赘述。
上文中结合附图从功能模块的角度描述了本申请实施例的装置1000。应理解，该功能模块可以通过硬件形式实现，也可以通过软件形式的指令实现，还可以通过硬件和软件模块组合实现。具体地，本申请实施例中的方法实施例的各步骤可以通过处理器中的硬件的集成逻辑电路和/或软件形式的指令完成，结合本申请实施例公开的方法的步骤可以直接体现为硬件译码处理器执行完成，或者用译码处理器中的硬件及软件模块组合执行完成。可选地，软件模块可以位于随机存储器，闪存、只读存储器、可编程只读存储器、电可擦写可编程存储器、寄存器等本领域的成熟的存储介质中。该存储介质位于存储器，处理器读取存储器中的信息，结合其硬件完成上述方法实施例中的步骤。
图11是本申请实施例提供的电子设备的示意性框图。
如图11所示，该电子设备可包括：
存储器1110和处理器1120，该存储器1110用于存储计算机程序，并将该程序代码传输给该处理器1120。换言之，该处理器1120可以从存储器1110中调用并运行计算机程序，以实现本申请实施例中的方法。
例如，该处理器1120可用于根据该计算机程序中的指令执行上述方法实施例。
在本申请的一些实施例中，该处理器1120可以包括但不限于：
通用处理器、数字信号处理器(Digital Signal Processor，DSP)、专用集成电路(Application Specific Integrated Circuit，ASIC)、现场可编程门阵列(Field Programmable Gate Array，FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件等等。
在本申请的一些实施例中，该存储器1110包括但不限于：
易失性存储器和/或非易失性存储器。其中，非易失性存储器可以是只读存储器(Read-Only Memory，ROM)、可编程只读存储器(Programmable ROM，PROM)、可擦除可编程只读存储器(Erasable PROM，EPROM)、电可擦除可编程只读存储器(Electrically EPROM，EEPROM)或闪存。易失性存储器可以是随机存取存储器(Random Access Memory，RAM)，其用作外部高速缓存。通过示例性但不是限制性说明，许多形式的RAM可用，例如静态随机存取存储器(Static RAM，SRAM)、动态随机存取存储器(Dynamic RAM，DRAM)、同步动态随机存取存储器(Synchronous DRAM，SDRAM)、双倍数据速率同步动态随机存取存储器(Double Data Rate SDRAM，DDR SDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM，ESDRAM)、同步连接动态随机存取存储器(synch link DRAM，SLDRAM)和直接内存总线随机存取存储器(Direct Rambus RAM，DR RAM)。
在本申请的一些实施例中，该计算机程序可以被分割成一个或多个模块，该一个或者多个模块被存储在该存储器1110中，并由该处理器1120执行，以完成本申请提供的方法。该一个或多个模块可以是能够完成特定功能的一系列计算机程序指令段，该指令段用于描述该计算机程序在该电子设备中的执行过程。
如图11所示，该电子设备还可包括：
收发器1130，该收发器1130可连接至该处理器1120或存储器1110。
其中，处理器1120可以控制该收发器1130与其他设备进行通信，具体地，可以向其他设备发送信息或数据，或接收其他设备发送的信息或数据。收发器1130可以包括发射机 和接收机。收发器1130还可以进一步包括天线，天线的数量可以为一个或多个。
应当理解，该电子设备中的各个组件通过总线系统相连，其中，总线系统除包括数据总线之外，还包括电源总线、控制总线和状态信号总线。
本申请还提供了一种计算机存储介质，其上存储有计算机程序，该计算机程序被计算机执行时使得该计算机能够执行上述方法实施例的方法。或者说，本申请实施例还提供一种包含指令的计算机程序产品，该指令被计算机执行时使得计算机执行上述方法实施例的方法。
当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行该计算机程序指令时，全部或部分地产生按照本申请实施例该的流程或功能。该计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。该计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，该计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线(digital subscriber line，DSL))或无线(例如红外、无线、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。该计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。该可用介质可以是磁性介质(例如，软盘、硬盘、磁带)、光介质(例如数字视频光盘(digital video disc，DVD))、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的模块及算法步骤，能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本申请的范围。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，该模块的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个模块或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，装置或模块的间接耦合或通信连接，可以是电性，机械或其它的形式。
作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理模块，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。例如，在本申请各个实施例中的各功能模块可以集成在一个处理模块中，也可以是各个模块单独物理存在，也可以两个或两个以上模块集成在一个模块中。
以上仅为本申请的具体实施方式，但本申请的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本申请揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本申请的保护范围之内。因此，本申请的保护范围应以该权利要求的保护范围为准。

Claims (14)

  1. 一种基于用户画像的数据处理方法，其特征在于，包括：
    获取多个用户数据以及所述多个用户数据中最后一个用户数据的生成时间；
    若从所述多个用户数据中每读取一个第一用户数据，则生成所述第一用户数据对应的用户特征；
    若所述第一用户数据的生成时间和所述最后一个用户数据的生成时间一致，则将所述第一用户数据对应的用户特征存储至目标数据库；
    若所述第一用户数据生成时间和所述最后一个用户数据的生成时间不一致，则不将所述第一用户数据对应的用户特征存储至所述目标数据库；
    基于所述目标数据库中的用户特征生成用户画像。
  2. 根据权利要求1所述的方法，其特征在于，还包括：
    获取第一用户特征查询请求；
    响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征；
    若所述目标数据库中未存储所述目标用户特征，且所述目标用户特征由所述目标数据库中的多个子用户特征构成，则将所述目标用户特征分解为所述多个子用户特征；
    基于所述多个子用户特征进行用户特征查询。
  3. 根据权利要求2所述的方法，其特征在于，所述响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征，包括：
    响应于所述第一用户特征查询请求，将所述第一用户特征查询请求转换为含义相同的第二用户特征查询请求；
    响应于所述第二用户特征查询请求，在所述目标数据库中查找是否存储有所述目标用户特征。
  4. 根据权利要求2或3所述的方法，其特征在于，所述响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征之前，还包括：
    对所述第一用户特征查询请求的发送方进行权限校验；
    所述响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征，包括：
    在对所述发送方进行权限校验通过时，响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有所述目标用户特征。
  5. 根据权利要求4所述的方法，其特征在于，所述对所述第一用户特征查询请求的发送方进行权限校验，包括：
    获取所述发送方的标识；
    根据所述发送方的标识确定所述发送方的权限范围；
    若所述发送方的权限范围包括查询所述目标用户特征的权限，则确定对所述发送方进行权限校验通过；
    若所述发送方的权限范围不包括查询所述目标用户特征的权限，则确定对所述发送方进行权限校验未通过。
  6. 根据权利要求2或3所述的方法，其特征在于，所述响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征之前，还包括：
    确定所述第一用户特征查询请求对应的查询接口；
    所述响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征，包括：
    基于所述第一用户特征查询请求对应的查询接口，响应于所述第一用户特征查询请求，在所述目标数据库中查找是否存储有目标用户特征。
  7. 根据权利要求6所述的方法，其特征在于，所述确定所述第一用户特征查询请求对应的查询接口，包括：
    若所述第一用户特征查询请求包括用户标识时，则确定所述第一用户特征查询请求对应的查询接口为联机事务处理OLTP接口；
    若所述第一用户特征查询请求不包括用户标识时，则确定所述第一用户特征查询请求对应的查询接口为联机分析处理OLAP接口。
  8. 根据权利要求1-3任一项所述的方法，其特征在于，还包括：
    判断所述目标数据库中是否存在异常的用户特征；
    若所述目标数据库中存在异常的用户特征，则生成提示信息；
    推送所述提示信息，以提示用户所述目标数据库中存在异常的用户特征。
  9. 根据权利要求8所述的方法，其特征在于，所述判断所述目标数据库中是否存在异常的用户特征，包括：
    获取所述目标数据库中在任一时刻存储的第一用户特征以及在所述任一时刻之前的预设时长内的至少一个第二用户特征；
    对所述至少一个第二用户特征进行统计，得到所述第一用户特征的分布范围；
    若所述第一用户特征不在所述分布范围内，则确定所述目标数据库中存在异常的用户特征；
    若所述第一用户特征在所述分布范围内，则确定所述目标数据库中不存在异常的用户特征。
  10. 根据权利要求1-3任一项所述的方法，其特征在于，所述若从所述多个用户数据中每读取一个第一用户数据，则生成所述第一用户数据对应的用户特征，包括：
    若从所述多个用户数据中每读取一个第一用户数据，则按照计算口径和应用场景，选择所述第一用户数据对应的用户特征的生成方法，所述生成方法是流生成方法或批量生成方法；
    基于所述第一用户数据对应的用户特征的生成方法生成所述第一用户数据对应的用户特征。
  11. 一种基于用户画像的数据处理装置，其特征在于，包括：
    第一获取模块，用于获取多个用户数据以及所述多个用户数据中最后一个用户数据的生成时间；
    第一生成模块，用于若从所述多个用户数据中每读取一个第一用户数据，则生成所述第一用户数据对应的用户特征；
    处理模块，用于：
    若所述第一用户数据的生成时间和所述最后一个用户数据的生成时间一致，则将所述第一用户数据对应的用户特征存储至目标数据库；
    若所述第一用户数据生成时间和所述最后一个用户数据的生成时间不一致，则不将所述第一用户数据对应的用户特征存储至所述目标数据库；
    第二生成模块，用于基于所述目标数据库中的用户特征生成用户画像。
  12. 一种电子设备，其特征在于，包括：
    处理器和存储器，所述存储器用于存储计算机程序，所述处理器用于调用并运行所述存储器中存储的计算机程序，以执行权利要求1-10中任一项所述的方法。
  13. 一种计算机可读存储介质，其特征在于，用于存储计算机程序，所述计算机程序使得计算机执行如权利要求1-10中任一项所述的方法。
  14. 一种包含指令的计算机程序产品，其特征在于，当所述计算机程序产品在电子设备上运行时，使得所述电子设备执行权利要求1-10中任一项所述的方法。
PCT/CN2022/107564 2022-07-25 2022-07-25 基于用户画像的数据处理方法、装置、设备、介质及程序 WO2024020708A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2022/107564 WO2024020708A1 (zh) 2022-07-25 2022-07-25 基于用户画像的数据处理方法、装置、设备、介质及程序
CN202280002410.2A CN115485676A (zh) 2022-07-25 2022-07-25 基于用户画像的数据处理方法、装置、设备、介质及程序

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/107564 WO2024020708A1 (zh) 2022-07-25 2022-07-25 基于用户画像的数据处理方法、装置、设备、介质及程序

Publications (1)

Publication Number Publication Date
WO2024020708A1 true WO2024020708A1 (zh) 2024-02-01

Family

ID=84395966

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/107564 WO2024020708A1 (zh) 2022-07-25 2022-07-25 基于用户画像的数据处理方法、装置、设备、介质及程序

Country Status (2)

Country Link
CN (1) CN115485676A (zh)
WO (1) WO2024020708A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112561559A (zh) * 2020-09-04 2021-03-26 上海东普信息科技有限公司 商户画像模型生成方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020161698A1 (en) * 2000-10-04 2002-10-31 Wical Kelly J. Caching system using timing queues based on last access times
CN111651471A (zh) * 2020-04-30 2020-09-11 中国平安财产保险股份有限公司 目标数据查询方法、装置、电子设备及存储介质
CN112417274A (zh) * 2020-11-17 2021-02-26 中国建设银行股份有限公司 一种消息推送方法、装置、电子设备及存储介质
CN113672401A (zh) * 2021-07-07 2021-11-19 浙江大华技术股份有限公司 一种批处理任务的触发方法、系统及计算机可读存储介质
CN114676161A (zh) * 2022-03-18 2022-06-28 中国建设银行股份有限公司 一种数据处理方法、装置、设备及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020161698A1 (en) * 2000-10-04 2002-10-31 Wical Kelly J. Caching system using timing queues based on last access times
CN111651471A (zh) * 2020-04-30 2020-09-11 中国平安财产保险股份有限公司 目标数据查询方法、装置、电子设备及存储介质
CN112417274A (zh) * 2020-11-17 2021-02-26 中国建设银行股份有限公司 一种消息推送方法、装置、电子设备及存储介质
CN113672401A (zh) * 2021-07-07 2021-11-19 浙江大华技术股份有限公司 一种批处理任务的触发方法、系统及计算机可读存储介质
CN114676161A (zh) * 2022-03-18 2022-06-28 中国建设银行股份有限公司 一种数据处理方法、装置、设备及存储介质

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112561559A (zh) * 2020-09-04 2021-03-26 上海东普信息科技有限公司 商户画像模型生成方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN115485676A (zh) 2022-12-16

Similar Documents

Publication Publication Date Title
US10853847B2 (en) Methods and systems for near real-time lookalike audience expansion in ads targeting
US20190080019A1 (en) Predicting Non-Observable Parameters for Digital Components
CN109299144B (zh) 一种数据处理方法、装置、系统及应用服务器
US10824608B2 (en) Feature generation and storage in a multi-tenant environment
US10182024B1 (en) Reallocating users in content sharing environments
JP2019536172A (ja) キー付きデータベースを使用して維持されるキー付きデータの検索および取得
CN113515545B (zh) 数据查询方法、装置、系统、电子设备以及存储介质
WO2014011866A1 (en) Predicting visibility of content items
CN111435376A (zh) 信息处理方法及系统、计算机系统和计算机可读存储介质
WO2024020708A1 (zh) 基于用户画像的数据处理方法、装置、设备、介质及程序
CN107291835B (zh) 一种搜索词的推荐方法和装置
CN116611411A (zh) 一种业务系统报表生成方法、装置、设备及存储介质
CN115757406A (zh) 数据存储方法、装置、电子设备及存储介质
US11356517B2 (en) Optimizing network utilization
US20210099545A1 (en) Optimizing network utilization
CN106817296B (zh) 信息推荐的测试方法、装置以及电子设备
US11354596B2 (en) Machine learning feature engineering
US20180240135A1 (en) Building Topic-Oriented Audiences
WO2023184831A1 (zh) 确定目标对象的方法及标识关联图的构建方法、装置
US20220230238A1 (en) System and method for assessing risk
CN106709353B (zh) 搜索引擎的安全性检测方法及装置
JP7223164B2 (ja) データインテグリティの最適化
CN111539208B (zh) 语句处理方法和装置、以及电子设备和可读存储介质
Drutsa et al. Prediction of video popularity in the absence of reliable data from video hosting services: utility of traces left by users on the web
US20230073643A1 (en) Predicting Events Based On Time Series Data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22952177

Country of ref document: EP

Kind code of ref document: A1