TOPK: A New Bipartite Ranking Algorithm for Enterprise Social Network

Abstract

Social networks bring us new perspectives and tools to rethink and solve the traditional problems in the enterprise business fields. Enterprise social network is presented to utilize the main factors in business comprehensively, like enterprises, employees and products. Based on the enterprise social network model we present TOPK algorithm to rank manufacturers and suppliers at the same time, which is improved on PageRank and HITS. Then we give the theoretical effectiveness analysis and convergence analysis. In addition we propose a method to recommend potential enterprise partners using TOPK to improve the accuracy. Then we perform the experiments on the dataset of China automobile supply network. Two evaluation methods are introduced to evaluate the results of enterprise rank. One method is based on the reliable public enterprise rank data, and another is to compare the results of clustering according to the enterprise rank. Through the two methods, our experiment results indicate that TOPK is better than the other algorithms at ranking enterprises. And enterprise rank from TOPK reflects more on car quality than the supplier amount of manufacturers.