A Study On Content Targeted Advertising Algorithm Based On Geographic Scopes

Precision is the direction of future development of online advertising, which requires researchers to further improve the advertising algorithms, and to fractionize the market. It is found by study that the more interest users have on the content of advertisements, the more it can improve the user experience and improve click-through rate, thus promotes the consonance of the benefits of users, advertisers and search engine suppliers.The basic idea of content targeted advertising is to extract the similarity between the text and copy of the advertising, and then advertise the advertisements related to the subject of the web page. Therefore, content targeted advertising match the advertisement and web page with the best similarity, which gives it a wide range of application. However, the research into the classification of topic by content targeted advertising is very few at the moment, which makes the classification of users very rare, the accuracy of content targeted advertising is quite low.Most of the products or services in advertising are sold at stores. Different chain stores at different locations could have different operating teams and marketing strategies. On the one hand, users may be only interested in certain products or services at certain locations. On the other hand, advertisers are more willing to highlight their advertising with location. Therefore, it is meaningful to consider geographic information in advertising.This article analyzes and concludes the related work, then provides a new contextual advertising algorithm based on extracting geographic information and the theme from web page. It is necessary for advertising algorithm to take the geographic information into account only when the text's theme has some relevance with the geographic information. So, we propose a new way to category webs into two classes which are relative class and not relative class. We propose a new way to assign a geographic scope for each text since it is possible that more than one place appears in one text, then we can calculate the distance between the geographic scopes of web and advertising, and this distance reflects the geographic similarity between certain web and advertising. In order to get a better precision when matching texts, we propose a new way to calculate the weights of key words based on homologies extension, which can also decrease the dimensions of key words vectors. There are too many webs and advertisings for a search engine to match, so it is important to improve the system's performance. We propose two models for filtering, and both of the two models are based on texts clustering and R-tree.Besides, this paper designs relative tests, which shows that the algorithm of content targeted advertising based on geographic information has a good recall rate and precision, and two filtering models can improve the performance of the advertising system.