Identifying Promising Technologies Considering Technology Convergence: A Patent-Based Machine-Learning Approach
提出融合距离、强度和多样性三个指标，用机器学习模型预测有前景技术，发现专利在技术融合中的中介作用对预测很重要。
With drastic changes in technology and its converging power in new product development, technology convergence has long been considered imperative in the innovation literature. Despite these efforts, previous articles neglected the importance of technology convergence in identifying promising technologies. To address this limitation, this article assumes that patents with high mediating power for subsequent technology convergence are likely to be promising. For this purpose, this article proposes the concept of convergence distance, which is measured by the differences in IPCs in backward and forward citations of patents, and defines it as the mediating power of technology convergence. Three indicators are defined: convergence distance, convergence intensity, and convergence diversity. Using these convergence-related indicators, we developed a machine-learning model to predict promising technologies. Consequently, the models with new evolution indicators outperformed the original models. Moreover, our suggested indicators turned out to be very important for predicting promising technologies, implying that the mediating power of technology convergence is very important for predicting future promising technologies and should be considered very significant for technology opportunity discovery.