The Design And Implementation Of Click Through Rating System Based On Contextual Advertising Platform

Contextual advertising is a kind of Internet advertising. It shows high related advertises on web pages based on the analysis of the page content and user intention. Contextual advertising is different from the traditional Internet advertising. Contextual advertising system mainly services small advertisers and long tail traffic, therefore, contextual advertising system has more advertises and more traffic. In the contextual advertising system, every retrieval advertises is a process of select advertises that is related to page content and user’s information. But due to performance reasons, we cannot use the most complex technology to calculate the relevance of each advertise in AD database. Therefore, contextual advertising system divides the retrieval process into two parts according to the relevance calculation complexity, which are advertising roughing selection and advertising sorting in the order. Advertising rough selection use smaller computational method to select some ads and then pass these ads to advertising sorting part to sort them. This paper focuses on advertising sorting stage, namely the system of click through rate.The traditional method of calculating relevance is base on cosine similarity to match page and the ad vectors. This method ignoring the advertisement click feedback is its mainly drawback. In this paper, we predict the click through rate of ad based on features of page, user and ad information. Our method uses more information than the traditional method. The most important is that we train a logistic regression model based on ad click feedback.This paper mainly introduces the design and implementation of the click through rate system. First, this paper introduces the present work on estimating the relevance among ad, page and user. Then this paper introduces the principle of click through rate system and then important technologies used in the system. In the following chapter, this paper analysis the system requirement base on the architecture and the design philosophy of contextual advertising system. According to the requirement this paper introduces the design and implementation of click through rate prediction system. Meanwhile it optimizes the system based on performance and the convenient of algorithm experiments. Finally the paper analyses the effect of the click through rate prediction system. The last part of the paper, we analyzed the prospect of this technology through the summary and outlook.