Searching the quantitative structure–property relationships (QSPR) is one of the most important tasks of the contemporary theoretical chemistry. The models obtained may be used for prognosis of the chemical substances properties on the basis of their structure and for searching compounds with predetermined properties. About 980,000 structure–property linear models have been constructed for parameter lgKa (Ka – acidity constant) of a number of organic compounds. Different training and test sets obtained by means of multiple random halving the initial set of compounds have been used to desigh these models. Molecular descriptors have been selected from some set of topological molecular parameters reflecting different particularities of molecular structures. The Wiener index, the generalized weighted connectivity index, the number of hydrogen atoms, some spectral characteristics of graph, representing molecule, etc. are among these parameters. About 90 the best models have been selected on the basis of some quantitative criterium, characterizing the model precision both on training and test sets. These models have been used for evaluation of lgKa for other compounds not included into the initial set of compounds, by calculation of lgKa for every model and averaging the results obtained. Besides, the structural particularities possessing the most significant influence on the given property have been derived on the basis of analysis of these models. The computer program Wolfram Mathematica 8 has been used in this work.