Content-Based AD Recommendation System For Mailbox

With the wide spread of internet and e-commerce, people are increasingly dependent on the Internet. People get news, music, video and other information from internet. The need for network information is constantly increasing. However, this situation makes it difficult for people to get access to effective information in front of so much mixed information. How to get information what users like has become a serious problem.This paper proposes the content-based ads recommendation system for mail. By analyzing the content-based recommendation algorithm and collaborative filtering recommendation algorithm, this paper determines the recommendation algorithm. Then this paper describes the overall design of the system’s framework.This system established a database and data dictionary. By parsing user’s past mails and log file, system got the keywords and stored them into the database. Also the system built data dictionaries. The system’s classification model is trained by SVM using latent semantic indexing. When a user receives an e-mail, this paper uses word vector tool to expand his keywords if his keywords are not enough. Then the system will classify the user’s interest based on the classification model. This paper uses a combination algorithm to achieve user’s ad recommendation results. This paper’s ads recommendation uses book as example.In this paper, content-based recommendation system is applied in mail platform. According to this feature, the system uses user’s e-mail content, operation records, contacts and other information as data source of content-based recommendation algorithm. At the same time, the recommendation algorithm of this paper is a combination of algorithms. The recommendation result combines using search engines and k-Nearest Neighbor algorithm, thus making the results that are presented to user more accurate.