The Design And Implementation Of Click-Through Rate Prediction System Based On Deep Learning

In recent years,with the rapid advancement of technology and Internet,online advertising has become one of the most profitable business models of Internet companies,drawing more and more attention of public and showing huge market potential.For example:search ads of search engines such as Google,Bing,Baidu,and recommended ads of e-commerce platform such as Amazon,eBay,Alibaba have brought huge benefits every year.Click through rate is a measure of the success of an online advertising campaign.Improve the accuracy of the estimated CTR can improve the user experience,enhance the effect of publicity advertising,and bring more advertising revenue.So how to predict the CTR more accurately has become a top priority.The industry mostly uses a large number of features extracted from the logs to build a logistic regression model and uses historical click data come training model approach.This method not only can't be a good solution to the new ads haven't historical click data,but also will occur the situation that the high dimension vector extract from the log will cause the computational complexity overload in the model training process.It will greatly affect the accuracy of the estimated online advertising CTR.Aiming the online advertising system of Dianping,this paper research CTR prediction system based on deep learning,the main work includes:1.Analysis the requirements on functionality and performance to the online ads CTR prediction system of Dianping.2.Complete the design and implementation of online ads CTR prediction system based on deep learning.3.Put the CTR prediction system in two test environment to see the effect and focus on data selection and feature selection has done a performance optimization...