The technique of segmenting data samples in order to improve indicators of the classifying algorithms quality is considered. It takes into account the factors that influence the change in the ranges of values of the target variables. Identifying impacts on current and anticipated situations allows for the segmentation of data samples. As a result, the ranges of the studied variables and outliers are reduced, and noisy data is removed. An experiment performed using a split sample is described. The results of the assessment were obtained for each classifier on the general sample and on the segments.