| Academic networks derived from research papers, in particular citation and co-author networks, have been studied widely. Although networks in an individual discipline particularly physics have been studied substantially, the difference across different disciplines remains unclear. This thesis shows that networks generated in computer science differ greatly the networks in Physics. The data used in our experiment contain more than two million papers in DBLP and half a million papers in Physical Review journals. We observe that both citation networks can be classified as scale-free networks. Papers in DBLP has a shorter life than PR. And physicists collaborate more closely than computer scientists in both citation and co-author networks. Collaborations evolve over time in both disciplines. For the ranking of papers, we find that the traditional PageRank algorithm is not appropriate for citation networks. We investigate the small-world characteristics in both kinds of networks in terms of small average shortest path. |
