Author ranking based on personalized PageRank

• Using publication citation networks gives better author orders than author networks.
• The uniform distribution of publication values among authors gives the best results.
• PageRank enriched by journal values is the best of our author-ranking methods.
• PageRank-based methods outperform all tested non-iterative approaches.
• H-index has low resolution and provides worse author ranking than publication count.

In this paper we evaluate citation networks of authors, publications and journals, constructed from the ISI Web of Science database (Computer Science categories). Our aim was to find a method with which to rank authors of scientific papers so that the most important occupy the top positions. We utilized a hand-made list of authors, each of whom have received an ACM Fellowship or have been awarded by an ACM SIG (Artificial Intelligence or Hardware categories). The developed method also included the adoption of the PageRank algorithm, which can be considered a measure of prestige, as well as other measures of significance (h-index, publication count, citation count, publication's author count), with these measures analyzed regarding their influence on the final rankings.Our main objective, to determine whether a better author ranking can be obtained using journal values, was achieved. The best of our author ranking systems was obtained by using journal impact values in PageRank, which was applied to a citation network of publications. The effectiveness of the ranking system was confirmed after calculations were carried out involving authors who were awarded after the final year used in our dataset or who were awarded in selected categories.