Data Filtering And Recommendation Based On Social Network

With the rapid development of network technology and the constant improvement of the ability of computer simulating human thinking, all kinds of industrial and daily applications, accumulated massive amounts of historical data. Humans have already transited from the famous era which is lack of information to the era of information overload. Faced with the huge, unordered Internet space, the traditional way of information searching already cannot meet the increasing demand of information for human. How to use the computer to find out potential and valuable information from the huge amounts of data quickly and efficiently, caused the earnest attention of people. At present, the development of the recommendation system can help users to locate the information who need conveniently and quickly. Analyzing the user’s interests is the key of the system. The emergence of social network provides a new comprehensive and abundant data set for us to further analyze user’s interest. Social Network becomes a hotspot among scholars at home and aboard now.In the past period of time, information processing technology became more and more mature. The text similarity computing has made great development. It has been widely used in the fields such as information retrieval, text classification, machine translation and so on. Some scholars made improvement with the method of text mining, eigenvalue extraction, similarity discriminant algorithm,and achieved some results. But most of these methods are for a specific application in the specific field only, whose applicable scope is small, especially in the field of Chinese information processing. Limited to the characteristics of Chinese language, the applicability of one algorithm has poorer performance in another field. New applications need to study a new algorithm to solve.This article is based on sina micro-blog. The method how to catch users’ interests using weibo and to make personalized recommendation to them is studied in this paper. Compared with the existing work, this paper mainly has the following different aspects. At first, in consideration of the shortcomings in the course of Chinese language information processing algorithm at present, extracting multiple attributes of text similarity criterion is proposed to applied to the micro blog. Through the multiple perspectives of feature extraction, improves the accuracy of text representation, and reduces the loss of semantic information in weibo text. Secondly, use external corpus to determine the user’s interest classes and to enrich the weibo semantics. What’s more, it is very helpful to overcome the problems that the number of user’s interest is difficult to determine which is caused by brief content of weibo. In addition, inspired by Ebbinghaus Forgetting Curve, we believe that human interest is not always the same. On the basis of the traditional recommendation algorithm, personalized recommendation algorithm based on time weight function is put forward, which is used to distinguish user’s real-time interest and to get rid of one’s overdue interest. Finally, we set several groups experiments. Compared with the traditional methods, the algorithm put forward in this paper can make a contribution to relieving the problems existed in the traditional methods and to improving the accuracy of recommendation effectively.