Scientific collaboration of post-Soviet countries: the effects of different network normalizations

The goal of the paper is to identify groups/clusters of countries with similar scientific collaboration “profiles” inside the group and to other groups of countries. The collaboration is described by co-authorship of a publication network that can be analyzed on the original co-authorship network. However, the network is dominated by large values in rows and columns of the most scientifically productive countries, which highly correlates with their sizes. This problem is especially relevant for countries with a big diversity such as post-Soviet countries. Normalization of the collaboration network allows making the countries’ collaboration comparable; therefore the question about its applicability and sensitivity to the collaboration structure is relevant. We analyze co-authorship networks of post-Soviet countries for the period 1993–2018. We use three types of network normalizations to make publication output of the countries comparable, namely affinity normalization, Jaccard normalization, and activity normalization. They provide different views on the scientific collaboration structure of the countries. We reveal the effect of the country size is the strongest when using the affinity normalization and it seems there is no countries ‘size effect for the activity normalization. Affinity normalizations reveal a big imbalance of collaboration between post-Soviet countries caused by their sizes. Russia has a great impact due to its size. Jaccard normalization reveals countries` collaboration is influenced by their neighborhood or by the size of national sciences. Activity normalization detects the research potential of a particular country. We also observe during the past twenty-five years the scientific collaboration has significantly changed, and the previously dominant position of Russia is decreasing. New groups of intense scientific collaboration have formed, affected by geographical neighborhood.