Exploring Technology Evolution Pathways Based on Link Prediction on Multiplex Network: Illustrated as CRISPR

Exploring technology evolution pathways is essential since one can capture the best opportunity in a particular domain. Researchers attempt to exploit the critical trajectory from a historical perspective; however, only some steps forward to forecasting the future direction. This study proposes a new research framework to make reasonable predictions. Based on patents retrieved from DII, we construct a multiplex network consisting of co-citation and semantic layers. Specifically, we utilize the citation relationships between patents and extract technology topics with the Combined Topic Model(CTM), a powerful topic recognition tool. Subsequently, we employ the link prediction method to obtain future links and assemble them into a new co-citation network. We get credible predictions of future evolution trends by analyzing topics. To validate our framework, we take CRISPR, an emerging technology in gene editing, as a case study. Our experiments show that link prediction performs well in detecting future co-citation links, and the semantic layer further improves the prediction accuracy. We finally summarize seven potential directions and validate our predictions.