Social tie mining in company networks

ABSTRACTDetecting competitive relationship plays a key role in a company's strategy formulation. Despite the importance, however, there are few automatic approaches to this problem, except several methods based on web data, which inevitably involve data noise. The challenging points of this problem are (1) to look for a set of data that can represent general information of the companies appropriately and (2) to figure out how underlying factors influence the competitive relationships between companies. In this paper, we choose patent records as data source and propose a novel approach to competitor detection problem. Specifically, we construct a patent network and employ an algorithm based on Random walk with restarts (RWR). Experimental results validate the of the algorithm. We further discuss to what extent the underlying factors influence the competitive relationship and finally demonstrate several representative results through case study.