Predicting High Impact Academic Papers Using Citation Network Features

摘要

Predicting future high impact academic papers is of benefit to a range of stakeholders, including governments, universities, academics, and investors. Being able to predict 'the next big thing' allows the allocation of resources to fields where these rapid developments are occurring. This paper develops a new method for predicting a paper's future impact using features of the paper's neighbourhood in the citation network, including measures of interdisciplinarity. Predictors of high impact papers include high early citation counts of the paper, high citation counts by the paper, citations of and by highly cited papers, and interdisciplinary citations of the paper and of papers that cite it. The Scopus database, consisting of over 24 million publication records from 1996-2010 across a wide range of disciplines, is used to motivate and evaluate the methods presented.