Adaptation of mathematical methods and models of fractal analysis to the study of aggregated ecomic time series of insurance company data

Abstract

Purpose: this article offers a comprehensive approach to forecasting the insurance market based on the joint use of both classical and new «nonlinear» statistics. Discussion: the methods proposed and tested by the authors are presented as a multi-criteria (two-criteria) mathematical model for assessing the trend stability of insurance time series. As the first criterion, the authors proposed an indicator that reflects the memory depth of a time series in the form of a fuzzy set, obtained on the basis of R/S analysis, and the second criterion is the Hurst indicator. Results: using a two-criteria approach to assessing the trend stability of time series allows you to differentiate them by the indicator of trend stability and select working forecast models.