Click Rate Prediction Model Based On Deep Neural Network

In the era of big data,information overload has arisen in all aspects of our lives.It becomes more and more difficult for users to find information that they are interested in from a large amount of information,and for information producers,it is also becoming more and more difficult for them to make the information they produce stand out from the vast amount of information.Recommendation system is produced under such premise.The main purpose of recommendation system is to recommend the right information to the right person.For users,the recommendation system can push the content of interest to users.For businesses,recommendation system can provide personalized services to users and increase revenue.In the recommendation system,click-through rate estimation is a very important link.To judge whether a product is recommended or not,it is necessary to estimate the click-through rate based on the click-through rate.This thesis mainly studies the problem of click-through rate prediction model based on deep neural network,focusing on current hot and advanced deep learning technology.Aiming at the existing deep click-through rate prediction model,this thesis improves the existing deep click-through rate prediction model at the two levels of low-dimensional feature expression and combination feature expression from the expression of low-dimensional feature and combination feature expression.It mainly introduces the idea of "field" in FFM into deep click-through rate prediction model and attention mechanism into depth.Two aspects of the click-through rate prediction model.The main work and innovations of this thesis include:1.A click-through rate prediction model based on FFM depth neural network is proposed.The main idea of this method is to introduce the idea of "field" in FFM model into the deep click rate prediction model,and to express the low-dimensional input features in multiple dimensions through the idea of "field",so that the model can better discover the rules contained in each feature.Finally,the deep neural network technology is used to predict the click rate.Relevant work has been sorted out and published patents.2.A click-through rate prediction model based on attention mechanism of deep neural network is proposed.The main idea of this method is to introduce the attention mechanism which has achieved good results in the field of image and natural language processing into thedepth click rate prediction model.Attention mechanism can better express the interaction between features and features,so that the model can better express the rules between features,and ultimately improve the performance of the model.The two tasks of this thesis are to improve the existing depth click rate prediction model from the low-dimensional feature expression level and the combined feature expression level,so that the performance of the model can be improved under the same parameters.Therefore,this method has broad application prospects.