The Algorithm of Co-Authors Selection for Preparing Scientific Works based on Gaussian Models and Data from the E-library Web Resource

Abstract

The article deals with the urgent task of selecting co-authors of scientific works using cluster analysis methods. In particular, on the basis of the resources of the scientific electronic library in Russia, E-library, test data on scientists were selected. These data were used to unite scientists into clusters according to interests and topics of their publication.  To solve this problem, a clustering method based on Gaussian mixture models (GMM) was used. The result of the research groups selection showed that the algorithm is able to qualitatively select scientists with common interests. To assess the effectiveness of the algorithm, the clustering results were checked, where the groups of scientists who already had common publications were chosen as the base. The obtained clustering accuracy was 100\% according expert assessment and exceeded the indicators obtained using the K-means algorithm.