Research And Implementation Of Targeting Engine Based On SVM And Hilbert R-tree

Along with the rapid development of Internet, There are various kinds of information on the Internet. However, people are interested in different information. while people get information which they are interested in from Internet,they are troubled by worthless information every day. How to accurately push information to users is a pressing problem for researchers.This paper mainly studies the related technologies in achieving an accurate real-time targeting system, it mainly include: data mining technologies、distributed storage-related technologies、distributed computing-related technologies,、spatial indexing algorithms,、distributed messaging systems and other related technologies。On this basis, it designs and implements a real-time targeting service framework that supports multi-dimensional user attributes. this paper studies are mainly as follows:(1)Summarized the development status, current research and practical significance of targeting system, and the related knowledge about targeting system, and points the features of efficient real-time targeting system.(2)The paper studies the principles and methods of data mining technology based on user information, and designs and implements a SVM classifier to classify the user’s attributes.(3)The paper studies the features of user’s attributes and classify user’s attributes. On this basis,it designs and implements corresponding user’s attributes subsystems using the idea of dividing and conquering.(4)The paper studies the related technologies of distributed key/value storage engine,and improves the open source distributed key/value storage engine Tair which basesd on consistent hashing algorithm.(5) The paper studies current various types of spatial indexing algorithm,and designs and implements the location subsystem basesd on Hilbert R-tree.Finally, we design and implement the targeting engine based on the above techniques, and does the functional and performance experiments.The results of tests show that the targeting engine has outstanding performance in terms of accuracy and real time compared with the current targeting engine used in Internet industry.