Research On The Prediction Of Advertising Conversion Rate Of Mobile APP

In today's Internet era,the use of smart phones is becoming more and more popular,mobile app advertising is favored by many people because of its mobile,interactive and wide audience characteristics,and has become an important branch of Internet advertising.Generally speaking,the effect of advertising is measured by the click through rate,but for advertisers,the conversion rate has a more direct relationship with their earnings.Therefore,the research on the prediction of the conversion rate of advertising has a strong practical value.However,due to the high dimension of the transformation data and the high sparsity of the transformation data,the current research results on the prediction of advertising conversion rate are not ideal.According to the background described above,through the modeling of users,advertisements and context information,this paper combines traditional machine learning algorithm and deep learning algorithm into the advertising conversion rate prediction experiment to improve the prediction accuracy of conversion rate.The specific contents are as described below:First of all,a series of operations,such as data cleaning and feature processing,are carried out on the advertising conversion log data and external feature files.The Logistic Regression model,the factorizer model and the DeepFM model are respectively established to predict.Through the conclusion of the experiment,it can be knowed that the DeepFM model,which combines the Deep Neural Network and the Factorized Machine model,has the best prediction effect.In order to further improve the prediction effect of the model and improve the model,mainly to mine the original feature data,a feature extraction means in view of users and app is raised,and the extracted features and the original data features are fused to get a new training set and test set,so as to build the above three models for prediction.From the results of the experiment,we can know that after the improvement of feature extraction,the prediction accuracy of the model is improved,and the DeepFM model is still the best one.