Technical note | Predicting Trends in Peer-Reviewed Publications

ABSTRACT

This work investigated a method for predicting future trends in peer-reviewed publication data related to science and technology. The developed method use non-linear regression to fit a particular type of s-shaped curve. The method's prediction accuracy is measured on historical publication data of mature technologies. It was found that the prediction accuracy is acceptable for short-term (5-10 years) predictions, but declines for long-term (10+ years) predictions. The method was also used to predict the future publication trends of novel technologies. It was found that the publication rate of various upcoming technologies is expected to grow. In particular, CRISPR and deep learning are expected to grow rapidly. Some technologies however, are already at the peak of their trend and are expected to decline in the future.

Executive Summary

The Concepts and Futures (C&F) group, within Joint and Operations Analysis Division (JOAD) is a collaborative research facility for the study of emerging and disruptive technologies. The team aims to identify areas of threat and opportunity in developing technologies and provides foresight to policy, strategy and capability development for the Australian Defence Organisation (ADO) and its strategic partners.

In support of the technology futures research, this work investigated a method for predicting future trends in peer-reviewed publication data related to science and technology. The developed method uses non-linear regression to fit a particular type of s-shaped curve, called the Gompertz function. The method’s prediction accuracy was measured on historical publication data of mature technologies. It was found that the prediction accuracy is acceptable for short-term (5-10 years) predictions, but declines for long-term (10+ years) predictions. The method was also used to predict the future publication trends of novel technologies. It was found that the publication rate of various upcoming technologies is expected to grow. In particular, clustered regularly interspaced short palindromic repeats (CRISPR) and deep learning are expected to grow rapidly. However some technologies, such as metal foam and organic light-emitting diode (OLED), have already reached the peak of their trend and are expected to decline in the future.