Prediction Of Click-through Rate Of Internet Advertising Based On Genetic Neural Network

In the 21 st century,because of the huge increasing development of computer science and information communication technology,gives the lives of human important impact and the Internet industry has made rapid and substantial progress.At the same time,all kinds of real industries began to be closely combined with the Internet,and Internet advertising as a new form of advertising came into being in this case.Internet advertising has great commercial value,because it can bring huge profits to Internet advertising platform,advertisers who produce advertisements and e-commerce companies who need to put advertisements.The Internet click through rate(CTR)can be used to directly measure the popularity of advertising,directly show the number of times watched by users,and the revenue obtained by the advertising platform is the product of single click through rate and click through rate.Therefore,it is beneficial for advertising media,Internet platform and e-commerce companies to accurately estimate the click through rate of advertising.Chinese styleAfter summarizing the research results of CTR prediction,this paper makes a systematic study on the related concepts and models of Internet advertising,which will be helpful for the follow-up theoretical research and practical application;secondly,this paper analyzes the problems related to CTR and integrates the relevant characteristics and attributes Finally,this paper applies genetic algorithm and neural network to the prediction of Internet advertising click through rate,and tests its effect.This paper mainly does the following four aspects of work:(1)This paper analyzes the prediction problem of Internet ad click through rate in detail and puts forward the process of solving the prediction problem.Firstly,it summarizes the general characteristics of Internet advertising click through rate data,and then divides it into classification problems according to the characteristics of advertising click through rate problem.Finally,neural network and genetic algorithm are selected as prediction models through model selection,and solutions are given.(2)The characteristic engineering processing of relevant data is carried out.Firstly,the missing value processing,cross analysis,feature fusion and coding are done for the data set.Finally,PCA is used to reduce the dimension of the encoded data,and 11 dimension data are obtained.(3)A prediction model of Internet advertising click through rate is established.First,the prediction model of advertising click through rate of genetic neural network is established,and then the data processed by feature engineering is combined with random forest and gradient lifting tree(GBDT)model to get new feature data.Finally,it is substituted into the genetic neural network model to get the accuracy and AUC value of this model.(4)According to the experimental data set,the click through rate of models such as logistic regression,decision tree and random forest are predicted,and the experimental results are compared with the genetic neural network.It is concluded that genetic algorithm combined with neural network can amazingly make the prediction accuracy of Internet advertising click through rate get higher.