Advertising Recommendation Model Based On Graph Embedding

With the popularity of the Internet,information based on the Internet has appeared in large numbers.Although this information can meet more needs of users,due to the explosive growth of information,users want to obtain the information they want in massive amounts of data.It is increasingly difficult,that is,although the amount of information has increased,the utilization of information has been greatly reduced,which has led to the overload of information.The recommendation system is a common means to solve the problem of information overload,and it is often responsible for information filtering.It is necessary to select users interested in a large amount of information.The recommendation system is a personalized system that can recommend the information and products of interest to the user according to the characteristics of the user,the attributes of the information,the history of the user.Different from the search engine,the recommendation system can personalize the user's behavior through the user's behavior,and select the content that the user is interested in from the plurality of candidate sets,thereby reducing the user's search time and increasing the utilization of the information.Personalized referral services are the primary role of the recommendation system,and a good recommendation system can make users dependent on it.The advertising recommendation system is an important branch of the recommendation system.In recent years,academic researchers have continuously proposed models specifically for advertising recommendation scenarios.In the industrial world,because advertising recommendations can generate a lot of profits,it has also received widespread attention.The advertisement recommendation systemrecommends the advertisements of interest to the user by analyzing the interests of the users and the characteristics of the advertisements.The paper first introduces the related concepts and research status of the advertising recommendation system,and then introduces the commonly used recommendation models.Then a deep learning advertising recommendation model based on graph embedding is proposed.Firstly,the image with the advertisement as the node is constructed by the user's interest in the advertisement.Then,the nodes in the graph are mapped into low-dimensional vectors by the graph embedding algorithm,and the output of the graph embedding is used as the input of the deep learning model.Finally,the other mainstream recommendation models are compared by experiments.The experimental results show that the deep learning model based on graph embedding performs best in the advertising recommendation scene,and the effects of different parameters on the model are compared by experiments.The improvement of all indicators in the experiment also proves.The effectiveness of the graph embedding algorithm for the ad recommendation system.Finally,the depth learning model embedded in the graph is further optimized,and the attention mechanism is introduced to perform weighted averaging of the output embedded by the average aggregated graph by adaptive learning weights.The algorithm introduces the attention mechanism again after the feature intersection to weight-average the vector generated by the feature intersection.The attention mechanism enables the model to give a higher weight to those parts with greater classification.The attention mechanism is realized by the neural network.In order to reduce the over-fitting risk brought by it,the algorithm adds a regularization method to the attention mechanism to solve the over-fitting problem.Finally,through experimental comparison,it is found that the convergence speed of the model is faster after adding the attention mechanism.Compared with the previous model,allthe indicators have a certain improvement.Through experimental comparison,the attention mechanism is sensitive to the parameters.