Design And Realization Of DMP System In Mobile Advertising Real Time Bidding System

The rapid development of Internet technology and its applications has brought about tremendous changes in the advertising industry.In recent years,there has been a personalized advertising delivery method for target groups,which allows advertisers to purchase target audiences rather than fixed advertising outlets to increase the effectiveness of advertisements and reduce the advertisers' operating costs.This form of personalized advertising is called RTB(Real Time Bidding),and its goal is to achieve "the right ads are presented to the right people in the right way at the right time." In the process of RTB ad delivery,the system for extracting,storing,and searching for the target population is called DMP(Data Management Platform),which is the key to implementing RTB.To meet this demand,this paper designs and implements a server-side DMP system iServer that supports high concurrency,high availability,and horizontal expansion.This main research contents are:(1)A DMP system that uses multi-dimensional user data composed of user interest,geographical location of mobile devices,and user spending habits,for the DSP(Demand Side Platform demand side platform)to produce the "most appropriate" advertising content.(2)The Redis-based iServer data storage scheme is adopted,and a distributed processing strategy is adopted to meet the requirements of high availability and DMP high concurrency with a single response within 5ms.(3)Adopting the data acquisition scheme based on Apache Kafka's distributed publish and subscribe message system,the user's IP location data of the mobile device and the third party user's interest data can be reported to the iServer in real time to meet the real-time requirements of the iServer.(4)A user interest data extraction method based on page keyword extraction technology is adopted.Through comprehensive integration of user web page browsing behavior,a general web page body extraction algorithm based on a row block distribution function is used for text extraction and reverse maximum matching.The algorithm performs word segmentation and finally uses a weighting algorithm based on TF-IDF to extract keywords that can highly summarize the content of the page,and then weights and normalizes the keyword of all pages of the user to quantify each keyword.The degree of importance of the user to form user interest data.(5)Aiming at the lack of unified cookie standard in mobile APP and not uniquely identifying the user's unique method,a mobile DMP implementation method for distinguishing users based on real-time IP is proposed.(6)Based on the above key technologies,the iServer prototype system was designed and implemented.It has the advantages of high concurrency,low latency,high availability,and horizontal scalability.Through the functional test and performance test of the iServer key module,the feasibility and effectiveness of the solution of this paper are confirmed.