Forecasting The New Advertising Click Rate Based On Bayesian Network

With the rapid development of the Internet in recent years, the Internet advertising industry has become a new market area is different with the traditional advertising industry. In this area, the Computational Advertising technology can realize the advertisement precise directional delivery plays the core position. The prediction of CTR (Click-Through Rate) provides the basis for the realization of advertisement precise directional delivery, but also maximizes the revenue of advertising media and improves user satisfaction. Thus, the prediction of CTR has certain research significance and practical value. To predict an ad’s CTR, historical click information is frequently concerned. However, to accurately predict the CTR of the new ads is challenging and critical for real world applications, since we do not have plentiful historical data about these ads.In this thesis, the similarity of ads can be inferred and discovered from the keywords of ads. Based on the similarity of ads, we could predict the CTRs of new ads by using the CTRs of the ads that are similar with the new ads. Adopting Bayesian network (BN) as the effective framework for representing and inferring dependencies and uncertainties among variables, in this paper, we establish a BN-based model to predict the CTRs of new ads. We build a keywords Bayesian network (KBN) through the analysis of the keywords data to reflect the direct similarity of keywords and the uncertainty of the similarity relation, and then we make use of the KBN inference to find the indirect similarity relationships between the keywords of the new ads. Based on the results of KBN inference, we can obtain the set of similar ads described by the similar keywords. Consequently, we can predict the CTR of the new ad by using the similar ads’CTRs that are known already.The main work and contributions of this thesis can be summarized as follows:■We built a Bayesian network of the keywords that are used to describe the ads in a certain domain, called keyword BN and abbreviated as KBN. We gave out an algorithm by performing statistical calculations through the keywords in order to construct the network structure of BN, and then found the direct similar keywords. ■We proposed an algorithm based on the Gibbs sampling for KBN’s approximate inferences to find similar keywords with those that describe the new ads■Based on the similar keywords, we obtain the similar ads and then calculate the CTR of the new ad by using the CTRs of the ads that are similar with the new ad. Experimental results show the efficiency and accuracy of our method.