The Design And Implementation Of Display Advertising System For Ride-hailing

With the rapid expansion of the mobile Internet advertising market,precision marketing has become the focus of research on advertising services.The existing advertising system cannot adapt well to the ride-hailing scene.In response to this phenomenon,this thesis studies the characteristics of ride-hailing Terminal advertising system.The system uses the CTR prediction model based on logistic regression to recommend display advertisements to online users,uses user history behavior logs to periodically train iteratively,and outputs advertisement sets in descending order.In order to accomplish the whole process of display advertising from configuration to terminal display,this system uses a layered architecture,based on the LNMP(Linux+Nginx+My SQL+PHP)website development environment,detailed design and implementation of terminal management,log collection,and display advertising.In the display advertising module,the configuration platform is used to configure advertisement channels,materials,and rules;the entity pool uses asynchronous scripts to obtain the advertising configuration to generate advertising entities;the delivery engine supports the delivery of millions of advertisement entities;the display controller reads the terminal Real-time advertisement information and user behavior data match the display rules to control the frequency of advertisement exposure on the terminal.Based on the historical data set of online ride-hailing users,the model is trained offline.Compared with the real value,the online recommendation accuracy rate of the model reaches 75.4%.After 21 days of running in real environment,the system functions normally,with satisfactory performance and favorable effectiveness in practical applications.