Inter-regional gaps in the investment sphere and sanction pressure on the national economy have actualized the issue of improving the efficiency of managing investment attractiveness at the mesoscale using modern high-precision methods of economic and mathematical modeling. A number of well-known thematic ratings are used to assess the investment attractiveness of the Russian regions, whose critical analysis made it possible to identify their main shortcomings. Therefore, in the course of the study, the authors make an attempt to build an adequate rating of the investment attractiveness of the constituent entities of the Russian Federation using artificial intelligence. The results of the retrospective assessment are deepened by clustering the Russian regions based on the achieved level of investment attractiveness by the method of Kohonen self-organizing cards. To implement the prognostic function, a Bayesian ensemble of dynamic neural network models was formed. As a result of the empirical study three hypotheses have been confirmed: on significant inter-regional gaps in the sphere of investment attractiveness; on a negative change in the cluster structure (in the field of investment attractiveness) of the Russian regions in 2013-2018 due to sanctions pressure on the national economy; and on the persistence of significant inter-regional gaps (in the investment sphere) in the medium term (on the example of the leading regions). The results can be used to update the provisions of the economic policy of a number of constituent entities of the Russian Federation. Moreover, the authors’ approach might be applied for analyzing other countries in the context of the UN Sustainable Development Goals at the meso-level.