Potential index: Revealing the future impact of research topics based on current knowledge networks

As the volume of scientific publications has been growing at an increasingly rapid speed, it is important to identify prominent research trends for scientists and institutions. While a considerable number of researchers have attempted to map the current state of scientific research, more efforts should be made to reveal potentially influential research topics. In this study, we investigate the relationship between the scientific impact of a research topic and the structure of its knowledge network. A novel indicator, potential index, is proposed to model topic impact based on the structural information. It is an immediate indicator with two components: knowledge novelty and diversity, which are operationalized using the concepts of betweenness centrality and network entropy. The empirical results show that potential index serves as a good predictor of future topic impact, with a high R2 and positive correlation. Its superiority sustains when used as the input feature of regression models. Moreover, the proposed index achieves better results, and the differences between it and other features become more prominent as the model complexity increases. Quantitative and qualitative analysis on the topic evolution process is also conducted to explain the change in the proposed indicator. This study contributes to the research of scientific impact modeling by establishing an explicit relationship between the impact of topics and the knowledge structure, and is thus helpful in predicting the potential impact of research topics.