Research On The Prediction Of Click Rate Of App Advertising Based On Machine Learning Mixed Model

With the rapid development of mobile Internet technology and the extensive development of mobile commerce,the rapid development of mobile Internet advertising industry has been promoted.According to the Research Report of ERI consulting,the scale of online advertising market has reached 2902.7 yuan in 2016,and the scale of mobile Internet advertising accounts for more than 60%.With the growing popularity of smart mobile terminals represented by smartphones,it is becoming more and more popular for advertisers to use mobile App applications for Internet advertising.App ads include a variety of forms such as built-in video advertisement,start screen advertisement,banner ad,integral wall advertisement,information stream advertising and so on.Compared with traditional network advertising,it has stronger media performance,information interaction and push precision,and can achieve better marketing goals.How to measure the effect of App advertising and improve its Click-Through Rate(CTR)and Conversion Rate(CVR)is the focus of both advertisers and App operators.Advertising click rate and conversion rate are important indicators to measure the effectiveness of advertising.Among them,the rate of advertisement clicking refers to the probability that the user clicks after the advertisement is officially released,and the rate of conversion refers to the probability that the user will click on the actual consumer behavior after clicking on the advertisement.Compared with the click through rate,the conversion rate directly affects the business objectives and economic benefits of advertisers,which is the key basis for advertisers to pay for advertising.Through the analysis and prediction of the App advertising click rate,it can not only judge the market efficiency after the advertisement,but also can further clarify the effect of the advertising effect on the behavior of the network consumer,thus helping the advertisers to carry out the marketing activities better and realize the commercial goal.This paper focuses on the App advertising click rate prediction problem,and systematically expounds the domestic and foreign research status and the basic theory of advertising click rate.On this basis,the comprehensive use of Random Forest,Gradient Boosting Decision Tree,Online Stochastic gradient descent,Field-aware Factorization Machine four kinds of machine learning algorithms construct the App ad click rate comprehensive prediction model-RF+LGFV,and analyze the application principle of the model in detail,and then verify the model based on the actual design.By comparing the prediction results with the single machine learning algorithm,it is proved that the hybrid machine learning algorithm has the advantage in App advertisement click rate prediction.The contributions of this article are as follows:1.App advertising click rate prediction feature selection.App ad click rate prediction feature selection.In this paper,two machine learning algorithms,such as random forest and gradient lifting decision tree,are used to extract significant features from the initial feature set,and the combination of two feature sets is used as a new feature set training prediction model,and the unimportant features are filtered.By extracting features,it can not only simplify the dimension,improve the efficiency of model training,but also prevent the interference of the low quality characteristics to the accuracy of the model,and improve the prediction accuracy..2.App advertising click rate prediction model construction.In this paper,in view of the disadvantages of single machine learning algorithm and subjective feature selection,a comprehensive prediction model of App ad click rate(RF+LGFV)is proposed in the comprehensive application of four kinds of machine learning algorithms,random forest,gradient lifting decision tree,random gradient descent,field perception factor decomposer model,and based on the real data of Tencent social App advertisements verify the validity and accuracy of the model,which provides an effective basis for advertisers to increase the click through rate of App advertisements.