This paper aims at an application of the statistical theory of spectra to the classification of chemical compounds. It has been shown that the moments of the intensity distributions may be used as molecular descriptors. The new descriptors have been tested using spectra of nitriles and amides. The dependence of the accuracy of the classification on the number of moments (up to the twelveth order) is discussed using model spectra.
Bibliographical noteFunding Information:
Supported by Ministry of Education and Science, Grant No. 2P04A 07229.
- DFT methods
- Genetic algorithm
- Method of moments
- Molecular similarity
- Statistical spectroscopy