In this paper we present an empirical evaluation of various techniques for feature selection that are applicable for analysis of funding decisions - whether of not to award funding to a specific scientific project. Input data are a set of review forms (questionnaires), filled in by domain experts, with final decisions of the expert committee about project funding. The data was provided by the Russian Foundation for Basic Research 1 . We compared various techniques and show that is makes more sense to compose an ensemble. The main contributions include machine-learning based methodology for retrospective decision analysis in the field of science management; a framework for proposals review quality control; a cross-domain criteria ranking for scientific projects.
DOI: https://doi.org/10.1109/IS.2016.7737417
PDF at the IEEE Explore digital library: https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=7737417
Devyatkin D., Suvorov R., Tikhomirov I., Grigoriev O. Feature selection techniques for scientific projects funding criteria analysis // Intelligent Systems (IS), 2016 IEEE 8th International Conference on. – IEEE, 2016. – p 167-172