The Research On Knowledge Push Based On The Feature Extraction Of Microblog
With the sharp increase of network information,providing targeted information service to users has become a hot issue.This paper adopts PCA to extract user feature with the aid of sina microblog and calculate contents user concern generating a candidate document library,next we push the documents having the highest similarity to user.On the basis,we put forward a knowledge push algorithm based on NMF and the constructed nonnegative matrix element is very consistent with the requirement of obtaining information in matrix element must be non-negative.Then,we design the word frequency matrix and build feature vector of user.By computing the similarity between feature vectors,the documents that user may be interested are obtained.Experiments show that,by comparing the precision rate and recall rate,compared with the traditional push algorithm,the proposed algorithm make a big difference on push quality,verifying the effectiveness of the proposed algorithm.
User interest Nonnegative Matrix Factorization Similarity calculation Knowledge push
Huang Weidong Ouyang Yafei
Nanjing University of Posts and Telecommunications,Nanjing 210023,China
国际会议
南昌
英文
1-9
2013-09-26（万方平台首次上网日期，不代表论文的发表时间）