Design And Implementation Of Advertising Push System Based On Spark

Due to the development of information technology and Internet technology,advertisers have turned their attention from flyers,newspapers,televisions,etc.to networks,computers,and mobile phones.However,overwhelming advertising not only wastes a lot of time for users,but also seriously interferes with daily life,and can't meet the needs of users and reduce the user experience.Therefore,finding user needs,discovering the differences between users,and delivering different advertisements for different users is the right choice to maximize the benefits of advertisers,users,and system developers.At present,targeted advertising is mainly to analyze the content that the user is currently browsing,to match the advertisement,and then to advertise.This completely ignores the user's interest,the delivery efficiency is not high,and the income is not obvious.Based on the above questions,this article continues to improve the content-based advertising algorithm.The user data is used to model the user interest,and the interest model is combined with the page keywords currently browsed by the user to match the advertisement relevance and improve the accuracy of the advertisement delivery.In the design and implementation of the advertisement delivery system in this paper,how to correctly mine the user's interests and hobbies,and establish the user's interest model is the most important,which determines the accuracy of the advertisement delivery.Therefore,when interest modeling is performed,based on the extraction of user theme modeling,not only the user's historical data is subjected to topic extraction and feature analysis,but also the user's behavior data and the user's content data are used to find a user similar to the user.Set,establish user history model,user behavior interest model and user content interest model,combine the three models to construct user interest model,and improve the accuracy and comprehensiveness of user interest model.In addition,in the advertisement and user interest matching,taking into account the user's temporary needs,the user's current page keyword weight is introduced,and the user interest model and the keyword weight are combined to perform advertisement matching,thereby further increasing the accuracy of the advertisement.Finally,due to the large amount of system data,complex calculations,and diverse calculations.Therefore,the system uses the memory-based spark distributed computing framework to train offline data such as topic models and user interest models.The real-time data such as keyword extraction for user browsing content is calculated using Spark Streaming real-time computing framework.Related data storage uses distributed storage systems such as HDFS,HBASE,and Hive.Use the redis cache database to speed up data requests for front-end requests.In this paper,the optimized user interest modeling and advertising recommendation algorithm is applied to the specific system,and the overall framework of the system is given.The implementation details of the core modules of the system are described in detail.In terms of system verification,this paper analyzes and verifies the effectiveness of advertising.Experiments show that the improvement of the user interest model and the design of the recommendation method can better identify the user's needs,bring new sense to the user,and increase the user experience.