BBS opinion leader mining based on an improved PageRank algorithm using MapReduce
Lincheng Jiang, Bin Ge, Weidong Xiao, Mingze Gao · 2013
Opinion leaders in bulletin board systems (BBS) play an important role during the formation of public opinion. Opinion leader mining has a positive effect on us to grasp and guide public opinion. The paper designs and implements an opinion leader mining system based on an improved PageRank algorithm using MapReduce. The improved PageRank algorithm uses the method of sentiment analysis to define the weight of the link between users. The system has three main steps. Firstly, adopt web crawler to crawl BBS data and preprocess the data collected. Then construct an online social network with the replying relations between posts and comments mapped to relations between posters and comment authors. Finally, use the improved PageRank algorithm to rank users in BBS in the Hadoop cloud computing environment. Contrast experiments with the origin Pagerank algorithm show that the system is accurate, efficient and practical.