Developing a supervised learning model for anticipating potential technology convergence between technology topics

Abstract

Recognized as essential for firms, a systematic methodology for identifying potential technology convergence opportunities has been the focus of many studies attempting to predict the emergence of new convergences between technology classes. However, because technology classes cover a fairly broad range of technology properties, the practical implications of the identified opportunities are not clearly articulated. Motivated to address this problem, this study develops supervised learning models to predict potential technology convergence between semantically coherent technology topics using cluster likelihood, link possibility, and technological similarity as input features. Patent data is utilized to generate technology topics and explore technology convergence between them, and a case analysis on wearables is conducted. Wearables, at the forefront of innovation, offer a dynamic environment converging disciplines like electronics and health sciences, making them an ideal subject for validating the proposed approach practically. The case analysis leads to the development of six individual models and one voting classifier that ensembles them together. The models perform favorably based on traditional metrics and indicate feasible and rational opportunities across various technologies, affirming the validity of our approach. This study's primary contribution lies in enhancing firms' R&D planning capabilities, enabling them to seize opportunities within specific technology topics.