Based On Social Tagging Of Personalized Information Recommendation Service

With the normal users participating in the Internet architecture and the informationcreation, serious information overload is becoming a big issue. How to help usersorganize, manage and retrieve information, become a huge challenge. Traditionalinformation search and information filtering technology has some shortcomings, such asdo not have the intelligence, the distinction between users does not high. By generating,adjusting the characteristics of user profile, recommended system use many differentalgorithms to make information resources actively, which is considered is an effectiveway to solve the information overload. With the participation of the majority of users,the Internet presents a social trend, its a typical representative of the social label, fromthe user point of view of resources, add annotations, reflecting the preferences, interestand concept space of the user’s, which are difficult to be obtained user information, sosocial tag are considered to be an efficient information organization.Tag-based recommendation system using the tag data which has links among users,resources and labels, to provide users with personalized recommendation services.There are three main ways of tag-based recommendation system: network-based method,tensor-method, and topic-method. Various methods solve problems of recommendedsystem to varying degrees, such as the network approach can solve the data sparsenessproblem, tensor methods can solve dimensionality reduction problem ofmulti-dimensional data, topic method has a reasonable model to explain. In this paper,from the network structure, tensor model and latent semantic layer points of view toanalyse tag-based recommendation system, proposed a hybrid recommendationframework, which can effectively combine network, tensor, topics methods with a verygood flexibility and scalability.Work of this paper includes:Proposed a hybrid tag-based personalized recommendation framework, combinethe output of different types of algorithms into a way to show users, resources, and thepossibility of tagging with four-tuples form, and eventually organizate and storage asranking cubic. Through a unified framework, tag-based hybrid personalized recommendation method to provide three recommendation services: recommendresources, recommend tags and users recommendation, and not require separateredesign for different recommendation requirements. Use Hadoop framework to achieveHOSVD, LDA and FolkRank three different types of algorithms.