The use of mathematical models is one of the main tasks in various areas of human activity for sustainable industry and life cycle analysis. The main method for this is the ordinary least squares. One of the possible benefits of analysis associated with the accuracy increasing for mathematical models in cases of non-stationary data trend is the heteroscedasticity adjustment. Heteroscedasticity occurs in case of changing the variance of the predicted variable for different intervals of the independent variable. First, heteroscedasticity was studied for economic problems, and it can also be considered the basis for sustainable economics. This paper is devoted to the development and analysis of methods for taking into account heteroscedasticity. In this case, five different models are studied. The first model is the geometric mean of the regression of Y on X and X on Y, calculated based on the ordinary least squares. The second and third models are based on weighted least squares and White’s method. The fourth model is new generalized moving average method. The fifth model uses an interval function to obtain heteroscedasticity equation. The sixth model is based on data clustering using the k-means method. The seventh model is based on a compressed correlation field. To determine the best model, their comparative analysis was carried out by calculating standard deviation and determination coefficient. The results of the study can be used to improve the accuracy of mathematical models in the case of heteroscedasticity.