Keyword: «statistical methods»

The purpose of this article is to find the best model for forecasting time series, taking into account the minimization of errors and high accuracy of the forecast. The method of comparative analysis of the most popular "traditional" econometric models ARIMA and SARIMA is used. Algorithms and models are implemented in the Python programming environment with the connected libraries Sklearn, Pandas, Numpy, and Statsmodels. As input data sets, we imported data on product sales for 5 years in 10 stores. The results of the study confirm the superiority of the SARIMA model, in which the RMSE error is 20% less than when using ARIMA. It is concluded that to improve the quality of the time series forecast, it is preferable to use an algorithm based on the SARIMA model.

The development of emergency situations on main pipelines running in the permafrost zone is affected by the loss of stability due to uneven freezing and thawing of rocks, the formation of heaving mounds, the displacement of the pipeline axis from the design position, the formation of ice along the route, pipeline rupture due to operating errors, defects equipment, corrosion, etc. The calculation of a risk score is often difficult due to the large number of uncertainties in the data used and available. The most effective and satisfactory assessment in terms of accuracy can be achieved with a combination of qualitative and quantitative methods – fuzzy models that combine qualitative expert knowledge and quantitative statistical methods.