Design And Implementation Of Advertising Click-through Rate Estimation System

As we all know,more and more Internet companies at home and abroad rely on advertising,such as Facebook,Google,and Yahoo,and Internet companies such as Ali and Baidu.They are all leaders in the Internet space,and the company’s main revenue is advertising.Especially in the rapid development of artificial intelligence in recent years,more and more companies are investing in the research of machine learning,neural networks and other related technologies.For example,these technologies improve the click-through rate and conversion rate of advertisements,and thus increase revenue.Compared with the previous advertising,the current network environment is undoubtedly more complicated and cannot be coped with without advanced technology as support.Therefore,how to filter out the ads that each user is satisfied with and clicks from the huge ad library is undoubtedly a huge challenge.In this complex background,this article considers how to design and implement a set of candidate advertisements that can be recalled from the advertising library for each user,and then accurately calculate the CTR(Click-Through)of each advertisement in the set of candidate advertisements according to the model Rate,click rate)estimation system.The advertising CTR estimation system needs to accurately use the basic information of the advertisement and the user,plus the user’s behavior information on the advertisement to calculate the probability of the advertisement being clicked by the user.Then use the click rate after the merge of multiple models as the basis for sorting the candidate advertisement set,select the advertisement with the highest click rate and recommend it to the user.This is the basic function that the system needs to realize.In order to achieve these functions,this paper designs and implements the advertising CTR estimation system based on collaborative filtering,LR(logistics regression)and FM(factorization machine)models.First of all,the background and significance of the research are introduced,and the current research status at home and abroad is deeply understood.Then,it introduces the related technologies such as model,feature engineering,and evaluation index in the system in detail.Then do a requirement analysis on the system,fully considering the functional and non-functional requirements of the system.Based on the results of the requirements analysis,the model is designed in outline,and the structure diagram of each module and the structure of the database are designed.According to the results of the outline design,each module is designed and implemented in detail,and each module is colluded.Finally,the system was tested accurately and experimentally.