Predicting future technological convergence patterns based on machine learning using link prediction

Technological convergence among different industries is an important source of innovation and economic growth. In this study, we propose a new framework for predicting patterns of technological convergence in two different industries. We first construct an inter-process communication co-occurrence network based on association rule mining. We then use a machine learning approach with various link prediction indices to predict future technological convergence patterns. Next, we use latent Dirichlet allocation (LDA) topic modeling to identify the keywords associated with technologies that are predicted to converge. We apply our proposed framework to a dataset of patents from the United States Patent and Trademark Office from 2012 to 2014 in the fields of chemical engineering and environmental technology. The empirical analysis results show that the prediction over a 4-year time interval using the random forest model achieves the highest performance. Moreover, the LDA topic modeling results indicate that the keywords "membrane," "air," "separation," "catalyst," "gas," "exhaust," and "particle" are descriptions of technologies that are likely to converge. This study is expected to contribute to technological and economic growth by predicting new technological fields that are likely to emerge in the future, and hence the directions that firms focusing on technological advancement should prepare for.