The Design And Implementation Of Ad Position Value Estimate Platform Based On Svm Regression

Advertising is one of the main profit-making ways of the many Internet sites. With the development of the Internet advertising industry, more and more Internet sites manage their ad positions by offering them to advertising alliance. Advertising alliance signs advertising plans with many advertisers and it will put ads on the most suitable sites by using a series of matching algorithms. Advertising generally charges in accordance with the clicks. Because the click-through rate of advertising is uncertain, the unit price of different ads varies. From the view of the website owners, for the reason that the profit of an ad position in the future is unknown, it will somehow affect the enthusiasm of the site owners. In this case, how to predict that how much income an ad position can obtain has become an urgent problem.This article describes a prediction method of advertising revenue. The main idea is to build a regression model by using a lot of historical data to make a support vector machine. The main work is divided into three parts. First, get the environmental characteristics data including the advertising environment characteristics and statistical data. The environmental characteristics data is obtained by crawling and analyzing the content of the advertising pages. Secondly, process data. After Analyze the correlation of all the characteristics data and the relevance between them, the data is filtered and converted according to need of the training model. Thirdly, train and optimize the model. Model is trained based on the existing support vector machine algorithm library. On this basis, model is optimized by using the parameter optimization, feature selection, the characteristics binding and such ways. This paper describes the work and achievements of these three stages. Finally it gives a brief summary and proposes some improvements in the future.