Online Advertising Prediction Model Based On Bayesian Method

Compared with the pattern of traditional media marketing, lots of companies choose searching engine advertising as the first choice to promote brand and reach the customers, for its low investment and high return soon.Therefore, it has become a hot research topic at home and abroad to improve the click rate and conversion rate of advertising. In order to solve this problem, the main ideas and results of this paper are as follows:First, the existing research on search engine advertising click-through rate hierarchical Bayesian model lacks effective processing advertising sparse and high dimensional data model, so the accuracy of the prediction results greatly reduced. This paper constructed an Ads click through rate prediction model based on LASSO variable selection method, which use a firm bid data to verify this model and the results show that the key factors that influence the Ads clicks is brand information, regional information and cost per click; it effectively overcome the existing models of Ads clicks in dealing with high dimensional and sparse data.Second, with the rise of online shopping, when advertisers deliver ads in the search engine, they will add links to download or buy the product, whose purpose is no longer a simple increase in Ads clicks,but to increase the conversion rate of products. Thus, forecasting the Ads conversion of Logistic regression model is put forward and the model consider the advertisement cost factors and nature factors that influence on the Ads conversion, establishing advertising cost model and advertisement nature factor model respectively. The example analysis shows that the Ads conversion mainly influenced by advertising nature factors of geographic information, trademark information and the specific product information, the results provide theoretical evidences for advertisers to develop search engine advertising investment strategies.Finally, based on the statistic conclusion, managers should pay attention to the impact of prior information on conversion rate for transformation of the existing records of advertisement, because it may mislead managers to developing an inappropriate strategy without taking the prior information into consideration. As a result, constructing the Ads conversion graph model and describe with the Bayesian method of graph model, we get the Bayesian model of the Ads conversion. With the actual transaction data:advertising nature factors on the conversion rate is greater than the influence of the advertisement cost factors. At the same time, by comparing the real data and the forecasting results of using the Bayesian model, model prediction accuracy is 65.81%.