Exploring technology opportunities by visualizing patent information based on generative topographic mapping and link prediction

Abstract The shortening lifetime of technology requires companies to make intensive efforts to continuously explore new technology. Although many researchers have proposed visualization methods to find technology opportunities, little attention has been paid to present detailed directions of technology development with specified characteristics of technology. Thus, this research aims to suggest a systematic approach to conducting technology opportunity analysis by visualizing patent information, such as patent documents and citation relationships. First, keywords that explain core concepts, functions, and so on are extracted from collected patent documents by text mining. Second, patents are visualized in a two-dimensional space, and vacant cells are identified with their estimated keyword vectors by generative topographic mapping (GTM). Third, since many vacant cells will be potential candidates for developing new technologies, link prediction tools can choose promising vacant cells to connect existing cells with potential, but not yet existent, cells. Finally, the results of prediction are tested by comparing the predicted cells with the actual developed cells. The research reported in this paper is based in three technologies that have emerging, stable, and declining patterns, in order to illustrate the proposed approach, and investigate in which types it is relevant. It is found that the proposed approach provided a good prediction performance in the case of a technology that has a stable pattern. In addition, among link prediction methods, a semantic similarity-based approach showed better prediction results than a machine learning technique due to modest data availability for training. Thus, the results of this research can help R&D managers plan and evaluate R&D projects for technology development.