Research On Prediction Of Click Through Rate And Detection Of Fraud In Internet Advertising

With the popularization and development of Internet technology and E-commerce,online advertising began to rise and gradually developed into a new and more important advertising mode.Advertising Click-Through Rates prediction is an important part of the online advertising system.Under the given user,commodity and scenario,the algorithm is used to predict the click probability of the website users to the online advertising.It is not only beneficial for advertisers to promote the products,but also to ensure the good experience of users.Click fraud detection algorithm is an important monitoring algorithm of advertising platform,which can accurately identify whether the received click records are normal or fraudulent,and further ensure the effectiveness and quality of the click after advertising.This paper analyzes and summarizes the reasons and difficulties of the prediction of Click-Through Rates and the detection of click fraud,and summarizes the traditional model,related technologies and general response strategies.This paper proposes a deep Click-Through Rates prediction algorithm and a click fraud detection algorithm for advertising click fraud.The main contents of this paper include the following two aspects:(1)A deep CTR prediction algorithm based on feature optimization and FTRL.This paper describes a series of training research on CTR prediction model based on neural network,and summarizes the CTR prediction model based on traditional deep learning.In view of the shortcomings of the traditional deep learning algorithm,this paper proposes a Feature-Optimization-based FTRL-DCN model of CTR prediction,and makes comparative experiments and analysis on the i Pin You dataset.The experimental results show that the model in this paper has better improvement effect compared with the base model,and the comprehensive prediction effect is the best compared with the typical CTR prediction model,with robustness and portability.(2)Detection algorithm of click fraud based on LightGBM.This paper describes a series of click fraud detection research based on traditional model,and summarizes the click fraud detection model based on traditional machine learning.In view of the shortcomings of traditional machine learning click fraud detection algorithm,this paper proposes a click fraud detection model based on FO-LightGBM,and carries out comparative experiments and analysis on Talking Data Ad Tracking public dataset.The experimental results show that the feature optimization method proposed in this paper has better improvement effect,and the comprehensive prediction effect of this model is the best compared with various typical click fraud detection models.