Personalized PUSH pushing method

Abstract

The invention discloses a personalized PUSH pushing method. The personalized PUSH pushing method comprises the following operation steps that S1, a big data storage system rapidly collects and stores mass data; s2, the big data processing engine quickly calculates and stores the collected data; and S3, the UDF integrates the seamless bridging big data batch processing flow and the real-time reasoning service. The accuracy of personalized PUSH content distribution is improved, stable improvement of indexes is powerfully promoted, user portrait and material label construction is perfected, the correlation and distribution efficiency of content distribution are improved, a big data calculation engine is utilized, thousands of people and thousands of aspects of differentiated distribution and targeted marketing are performed on massive users, and through re-construction of the existing PUSH, the content distribution efficiency is improved. And the personalized push service is more systematic.

Description

Technical Field

[0001] The present invention relates to a personalized PUSH method. Background Art

[0002] In the era of mobile Internet, information overload has become a major problem faced by users, and personalized push technology, as a key means to solve this problem, is becoming increasingly important.

[0003] Existing personalized push notifications are mainly implemented by business personnel manually formulating rules, selecting target audiences, and distributing corresponding rule materials. The personalized distribution throughout the day is composed of a series of numerous rule tasks. However, this personalized push notification method has certain shortcomings. The push notification method is configured based on the experience of business personnel, the trial and error cost is high, and there is a certain lag in capturing user behavior, resulting in a certain delay in the push notification effect. User interests are also relatively single, and it fails to fully portray the user's complete portrait for targeted marketing. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the high trial and error cost of the personalized push method in the existing technology, a certain delay in the push effect, and the failure to fully portray the user's complete portrait and conduct targeted marketing, and to provide a personalized PUSH push method.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] The present invention provides a personalized PUSH push method, which includes the following steps:

[0007] S1. Big data storage system quickly collects and stores massive amounts of data;

[0008] S2, the big data processing engine quickly calculates and stores the collected data;

[0009] S3 and UDF integration seamlessly bridge big data batch processing processes and real-time inference services;

[0010] S4, online reasoning service forms accurate user profiles and customizes strategies based on the bridged big data;

[0011] S5. Based on user portraits and combined with intelligent recommendation algorithms, push content is tailored for each user.

[0012] Preferably, the big data storage system adopts the file storage service in the enterprise's internal cloud storage solution, and ensures model sharing of multiple servers of the deep model through shared storage.

[0013] In this technical solution, high availability, consistency, and cross-server accessibility of model files are ensured, providing a solid data infrastructure for the project.

[0014] Preferably, the big data processing engine uses integrated Apache Spark as the core computing engine.

[0015] In this technical solution, the big data processing engine performs efficient distributed processing on massive amounts of historical site data, significantly improving data processing speed and flexibility.

[0016] Preferably, the Apache Spark is a fast and versatile computing engine designed specifically for large-scale data processing, and is better suited for MapReduce algorithms that require iteration, such as data mining and machine learning.

[0017] Preferably, the UDF refers to a user-defined function, which is generally used in data processing systems and big data platforms, such as Hive, MaxCompute (formerly ODPS), Spark, etc., to enhance the platform's data processing capabilities and flexibility.

[0018] Preferably, the UDF integration is applied to a big data processing engine, and by developing and integrating UDF custom functions, a big data batch processing flow and real-time reasoning services are seamlessly bridged.

[0019] Preferably, the online reasoning service adopts an online reasoning service built on the Spring Cloud framework, focusing on the accurate construction of user portraits, dynamic capture of interest preferences and intelligent distribution of personalized content, forming accurate user portraits based on bridged big data, and providing users with a highly customized experience.

[0020] Preferably, the intelligent recommendation algorithm uses Apache Flink combined with Kafka to build a real-time data stream processing system to achieve real-time monitoring and capture of user behavior on the site, providing real-time and accurate data support for the implementation of personalized PUSH push strategies, ensuring that the pushed content is tailored to each individual and reaches them accurately.

[0021] This technical solution innovatively integrates online inference services with offline big data batch processing, achieving seamless integration and successful application for the first time in the industry. Leveraging the powerful capabilities of a distributed computing engine, combined with user-defined functions (UDFs), it not only enables rapid, large-scale processing of massive amounts of data, but also promotes the tight integration of big data, inference services, and deep learning models, improving overall processing efficiency and effectiveness.

[0022] Preferably, Apache Flink is an open source stream processing framework developed by the Apache Software Foundation. Its core is a distributed data stream engine written in Java and Scala. Flink executes arbitrary stream data programs in a data-parallel and pipelined manner. Flink's pipeline runtime system can execute batch and stream processing programs. Kafka is an open source stream processing platform developed by the Apache Software Foundation and written in Scala and Java. The purpose of Kafka is to unify online and offline message processing through Hadoop's parallel loading mechanism and to provide real-time messages through clusters.

[0023] Preferably, the specific process in step S5 is:

[0024] Step 1: Data pulling, pulling customized policy data;

[0025] Step 2: Policy priority, determine the priority of the pulled policy data;

[0026] Step 3: Strategy summary, summarizing strategies for different time periods;

[0027] Step 4: Deduplication of strategies: Deduplication of repeated strategies in the summarized strategies;

[0028] Step 5: Content sorting: sort the content in the strategy after deduplication;

[0029] Step 6: Content distribution, distribute the sorted content;

[0030] Step 7: Content receipt.

[0031] On the basis of conforming to the common sense in this field, the above-mentioned preferred conditions can be arbitrarily combined to obtain the preferred embodiments of the present invention.

[0032] The positive progress effect of the present invention is:

[0033] 1. This invention integrates offline batch processing, online inference services, and real-time processing tasks into a personalized PUSH system, achieving independent decoupling and efficient collaboration between each service module. Specifically, the batch processing module focuses on reading user information from historical HDFS data and dispatching a distributed computing engine for in-depth analysis; the online inference service accurately captures user profiles and behavioral characteristics to drive precise push strategies; and the real-time processing task captures user behavior within the site in real time, dynamically characterizing their immediate preferences and ensuring the timeliness of pushed content.

[0034] 2. The present invention improves the accuracy of personalized PUSH content distribution and effectively promotes the steady improvement of indicators.

[0035] 3. This invention improves the construction of user portraits and material tags, and improves the relevance and efficiency of content distribution.

[0036] 4. The present invention utilizes the computing engine of big data to provide differentiated distribution and targeted marketing to massive users.

[0037] 5. The present invention makes the personalized push service more systematic by re-structuring the existing PUSH. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Schematic diagram of a personalized PUSH push method according to an embodiment of the present invention.

[0039] Figure 2 Schematic diagram of the S4 process of personalized PUSH according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention is further described below by way of examples, but the present invention is not limited to the scope of the examples.

[0041] The personalized PUSH push method includes the following steps:

[0042] S1. Big data storage system quickly collects and stores massive amounts of data;

[0043] S2, the big data processing engine quickly calculates and stores the collected data;

[0044] S3 and UDF integration seamlessly bridge big data batch processing processes and real-time inference services;

[0045] S4, online reasoning service forms accurate user profiles and customizes strategies based on the bridged big data;

[0046] S5. Based on user portraits and combined with intelligent recommendation algorithms, push content is tailored for each user.

[0047] The big data storage system adopts the file storage service in the enterprise's internal cloud storage solution, and ensures the model sharing of multiple servers of the deep model through shared storage.

[0048] It ensures high availability, consistency, and cross-server accessibility of model files, providing a solid data infrastructure for the project.

[0049] The big data processing engine uses integrated Apache Spark as the core computing engine.

[0050] The big data processing engine performs efficient distributed processing on massive amounts of historical site data, significantly improving data processing speed and flexibility.

[0051] The Apache Spark is a fast and general computing engine designed for large-scale data processing, and is better suited for MapReduce algorithms that require iteration, such as data mining and machine learning.

[0052] The UDF refers to a user-defined function, which is generally used in data processing systems and big data platforms, such as Hive, MaxCompute (formerly ODPS), Spark, etc., to enhance the platform's data processing capabilities and flexibility.

[0053] The UDF integration is applied to the big data processing engine. By developing and integrating UDF custom functions, the big data batch processing process and the real-time reasoning service are seamlessly bridged.

[0054] The online reasoning service uses an online reasoning service built on the Spring Cloud framework, focusing on the accurate construction of user portraits, dynamic capture of interest preferences, and intelligent distribution of personalized content. It forms accurate user portraits based on bridged big data and provides users with a highly customized experience.

[0055] The intelligent recommendation algorithm uses Apache Flink combined with Kafka to build a real-time data stream processing system to achieve instant monitoring and capture of user behavior on the site, providing real-time and accurate data support for the implementation of personalized PUSH push strategies, ensuring that the pushed content is tailored to each individual and reaches them accurately.

[0056] This innovative fusion of online inference services and offline big data batch processing technology has achieved seamless integration and successful application for the first time in the industry. Leveraging the powerful capabilities of the distributed computing engine, combined with user-defined function (UDF) technology, it not only enables rapid and large-scale processing of massive amounts of data, but also promotes the tight integration of big data, inference services, and deep learning models, improving overall processing efficiency and effectiveness.

[0057] Apache Flink is an open source stream processing framework developed by the Apache Software Foundation. Its core is a distributed data flow engine written in Java and Scala. Flink executes arbitrary streaming data programs in a data-parallel and pipelined manner. Flink's pipeline runtime system can execute batch and stream processing programs. Kafka is an open source stream processing platform developed by the Apache Software Foundation and written in Scala and Java. Kafka's purpose is to unify online and offline message processing through Hadoop's parallel loading mechanism and to provide real-time messages through clusters.

[0058] The specific process in step S5 is as follows:

[0059] Step 1: Data pulling, pulling customized policy data;

[0060] Step 2: Policy priority, determine the priority of the pulled policy data;

[0061] Step 3: Strategy summary, summarizing strategies for different time periods;

[0062] Step 4: Deduplication of strategies: Deduplication of repeated strategies in the summarized strategies;

[0063] Step 5: Content sorting: sort the content in the strategy after deduplication;

[0064] Step 6: Content distribution, distribute the sorted content;

[0065] Step 7: Content receipt.

[0066] Although specific embodiments of the present invention have been described above, those skilled in the art will appreciate that these are merely illustrative and that the scope of the present invention is defined by the appended claims. Those skilled in the art may make various changes or modifications to these embodiments without departing from the principles and essence of the present invention, and such changes and modifications are intended to fall within the scope of the present invention.

Claims

1. Personalized PUSH push method, characterized by: The personalized PUSH push method includes the following steps: S1. Big data storage system collects and stores massive amounts of data quickly; S2, the big data processing engine quickly calculates and stores the collected data; S3 and UDF integration seamlessly bridge big data batch processing and real-time reasoning services; S4, online reasoning service forms accurate user portraits and customizes strategies based on bridged big data; S5. Based on user portraits and combined with intelligent recommendation algorithms, push content tailored for each user.

2. The personalized PUSH method according to claim 1, characterized in that: The big data storage system adopts the file storage service in the enterprise's internal cloud storage solution, and ensures the model sharing of multiple servers of the deep model through shared storage.

3. The personalized PUSH method according to claim 1, characterized in that: The big data processing engine uses integrated Apache Spark as the core computing engine.

4. The personalized PUSH method according to claim 3, characterized in that: The Apache Spark is a fast and general computing engine designed for large-scale data processing, and is better suited for MapReduce algorithms that require iteration, such as data mining and machine learning.

5. The personalized PUSH method according to claim 1, characterized in that: The UDF refers to a user-defined function, which is generally used in data processing systems and big data platforms, such as Hive, MaxCompute (formerly ODPS), Spark, etc., to enhance the data processing capabilities and flexibility of the platform.

6. The personalized PUSH method according to claim 5, characterized in that: The UDF integration is applied to the big data processing engine, and by developing and integrating UDF custom functions, the big data batch processing flow and the real-time reasoning service are seamlessly bridged.

7. The personalized PUSH method according to claim 1, characterized in that: The online reasoning service adopts the online reasoning service built on the Spring Cloud framework, focusing on the accurate construction of user portraits, dynamic capture of interest preferences and intelligent distribution of personalized content. It forms accurate user portraits based on bridged big data and provides users with a highly customized experience.

8. The personalized PUSH method according to claim 1, characterized in that: The intelligent recommendation algorithm uses Apache Flink combined with Kafka to build a real-time data stream processing system to achieve real-time monitoring and capture of user behavior on the site, providing real-time and accurate data support for the implementation of personalized PUSH push strategies, ensuring that the pushed content is tailored to each individual and reaches them accurately.

9. The personalized PUSH method according to claim 8, characterized in that: Apache Flink is an open source stream processing framework developed by the Apache Software Foundation. Its core is a distributed data stream engine written in Java and Scala. Flink executes any stream data program in a data parallel and pipeline manner. Flink's pipeline runtime system can execute batch and stream processing programs. Kafka is an open source stream processing platform developed by the Apache Software Foundation and written in Scala and Java. The purpose of Kafka is to unify online and offline message processing through Hadoop's parallel loading mechanism, and to provide real-time messages through clusters.

10. The personalized PUSH method according to claim 1, characterized in that: The specific process in step S5 is as follows: Step 1: Pull data to pull customized policy data; Step 2: Policy priority, determine the priority of the pulled policy data; Step 3: Strategy summary: summarize the strategies in different time periods; Step 4: Remove duplicate policies from the summarized policies. Step 5: Content sorting: sort the content in the strategy after deduplication; Step 6: Content distribution, distributing the sorted content; Step 7: Content receipt.

Images