K-Means-Gradient Lifting Algorithm For Screening Advertisements On Network Platform

The charging fee for the enterprise to advertise the enterprise is divided into two parts,one is based on the position and location of the occupied page of the advertisement;The other part is charging based on the click through rate of the ad.The amount of profit obtained by the network platform is closely related to the click of the advertisement.The higher the click rate,the greater the profit of the network platform.Therefore,the click rate of the advertisement becomes a key factor for the network platform to filter the advertisement,and an appropriate click rate should be established for the advertisement click rate.Forecast model.The existing forecasting models include logistic regression,random forest and gradient lifting models..But the effect of prediction is difficult to achieve the desired state.Firstly,1928 advertisement data are collected from data DC.Each piece of advertising data contains the advertiser,the advertiser industry,the promotion activity,the level of the creative idea,the presentation form of the advertisement idea,the advertisement width,the advertisement height,and the landing page jump.10 indicators of turn,landing page download and advertising voice video;fill the missing data with the average of 3 neighbor samples,and then use the Gini index to filter the characteristics of the advertising data for advertisers,advertisers,promotions,creative ideas The level,the creative form of the advertisement,the advertisement width and the advertisement height,the advertisement data of the seven characteristic indicators are used as the sample training set,and finally the K-means-gradient lifting algorithm is used to predict the probability of the advertisement click,and a probability is given to the probability.The predicted probability exceeds this threshold,which is regarded as the advertisement click rate is high.The advertisement category output is 1.The network platform can accept the advertisement.When the probability of predicting the click is less than the threshold,the advertisement click rate is low,the advertisement category output is 0,and the network platform does not.The ad can be accepted.This paper illustrates the effectiveness of K-means-gradient lifting algorithm.Using the training data set in this paper,to learn the logistic regression,random forest and gradient lifting classifiers respectively,and classify the classification results of these classifiers with the K-means-gradient lifting algorithm.The comparison results show that the prediction accuracy of K-means-gradient lifting algorithm is the highest.