DATA ANALYSIS USING FRACTAL GEOMETRY AND SELF-SIMILARITY METHODS

At the moment, the researchers of the market and economic indicators (cycles) scarcely use (due to the utter skepticism) for their calculations mathematical techniques to find fractal patterns (self-similarities) determining the movement of the studied indicators trend (or some state of the studied sector of the agricultural economy). These tools showed their efficiency in predicting the macroeconomic time series of performance indicators of some regional participants in the sugar sub-complex of the agro-industrial complex. Some elements of such patterns have proven themselves well in the construction of indicators of advanced development. They belong to the class of express methods of trend identification. In terms of efficiency and time expenditures, they are significantly superior to mathematical tools such as artificial neural networks, genetic algorithms, fuzzy logic methods, etc. The paper implements the search for stable price patterns in the history of price quotations similar to the current values. The idea is that any price pattern has taken place in the past: having this pattern properly identified, it is possible to predict to a high precision the behavior of any segment of the agro-industrial market. The author considered the forecasting methods belonging to the class of phase-fractal analysis and self-similarity methods. Besides, the author emphasizes the adaptation of such techniques when predicting the indicators of regional participants in the sugar sub-complex of the agro-industrial complex. Within the practical part of the work, the author applied the elements of phase-fractal analysis for the spurious response rejection. It allowed significantly decreasing the information noise in one dimension spectra. The results of applied calculations and practical implementation confirmed the possibility of using the tool in predicting the economic performance of large industrial enterprises of the sugar sub-complex. The results obtained for the described models allow performing multivariate calculations for the same indicators. The results of using phase-fractal analysis and self-similarity methods in forecasting tasks demonstrated the possibility of solving them and confirmed their practical significance.

phase-fractal analysis, self-similarity, self-affinity, reduction, J.M. Hurst Cyclic theory, intermarket analysis, time scale, real-time scale selection