COMPREHENSIVE METHODOLOGY OF SUPPORTING INVESTMENT DECISIONS

Background. The paper highlights the need to modify traditional statistical methods that are based on the assumption of normal distribution of quotes and do not take into account more complex dynamic characteristics of financial assets. The author proposes a new methodology that includes a binary approach to selecting assets in a portfolio, where the basis for making a decision is the feedback received from multidisciplinary forecasting methods. The purpose of the study is to improve the efficiency of investment decisions by developing a comprehensive methodology for supporting decision-making based on the transformation, combination and synthesis of statistical and spectral methods for forecasting time series. Materials and methods. The comprehensive methodology includes forecasting methods ARIMA/ARMA, ARIMA/GARCH and Fourier decomposition, modified by the author. To make decisions on this basis, a general model and its special cases was developed - modified random forest and Adaboost algorithms. Results. Validation of the models included in the methodology was carried out in comparison with the classical Markowitz model on four world indices for different periods. In the vast majority of cases, the proposed models showed a better result than the classical model. Conclusions. The integrated methodology for supporting investment decision-making is more flexible compared to existing ones due to adaptation to the nature of time series and allows for increased investment efficiency, which was shown during validation. In the future, author plans to test the methodology on the Russian market with calculation of the economic effect.