Find in Library

The article is devoted to questions of accumulated data usage to find out regularities by means of pattern recognition methods that allow predicting formation of not synthesized substances and estimating its properties. The formal task of computer-aided inorganic compounds design is stated. An approach to reproduce of missing data in learning samples for computer-aided inorganic compounds design is proposed. It is based on combination of linear regression and interpolation taking into consideration the problem domain – inorganic chemistry. The approach is more powerful than methods of missed data reproduction used currently in information-analytical system for inorganic compounds design running at IMET RAS.