Research On Ad Recommendation System Based On Ads CTR And Tag Recommendation Graph Model

With the rapid development of information technology and the rapid spreading of Internet, the network information data is not only showing explosively growth, but also the contents of data are becoming more and more abundant. This makes it more difficult for users to obtain the useful information from vast amounts of data in the Internet in which they are interested. The arrival of the recommendation system and tag recommendation system is largely helping to organize the information from network and enables the user to find out the information they are interested in more quickly.At first, this paper introduces three traditional recommendation methods about the recommendation system, they are collaborative filtering method, content-based recommendation method and hybrid recommendation method. Then it introduces the details of these three methods. Secondly, this paper describes the characteristics of the tag recommendation system and its graph model and a classical tag recommendation algorithm based on the graph model. Then, this paper introduces the concepts and methods of Internet advertising and the advertise recommendation. According to the characteristics of the tag recommendation graph model, it puts forward an adapted advertising recommendation algorithm based on the tag recommendation algorithm and Ads CTR. In order to better adapt to the advertising graph model, this paper proposes further improved algorithms. The first adapted method based on the advertising recommendation graph model is not only considering the influence from the click behavior of similar users, but also reducing the dimension of computation space by the change of matrix, it avoids the unnecessary node in graphical model participating in the process of sorting. The second method changes the graph model in order to turn the undirected weighted graph to directed weighted graph, after that, we can use advertising recommendation algorithm based on the brand new graph. The last method takes the advantage of path in graph to measure the proximity of the nodes and produces the ad recommendations for certain nodes. In the end, we execute experiments on the established ad recommendation platform, it turns out that the three methods indicate better result. Therefore, in the recommendation of ads, establishing the ad recommendation graph model based on the tag recommendation graph to find the relationship between the nodes in the graph is a viable approach to recommend ads.