David Artashevich Gertsekovich

This article provides the investment analysis, with the portfolio method being one of the frontmost instruments. The traditional portfolio theory implies that its essential criteria such as expected return and risk rate for all considered financial instruments are estimated from the historical data of equal length and are invariant throughout the usage period of the model. The authors suggest the approach which is a development of the Markowitz model. The point being, to estimate the expected return and the risk rate for each financial instrument in focus, an effective (own) learning sample is determined. The extremum seeking for the target function is performed through the method of full enumeration (cycle optimization) which provides the global extremum of the optimization criterion with restriction to the maximum permissible risk level. The sample size is tried upon two quality criteria considering accuracy of forecasts, namely: 1) the minimum condition of the sum of squared deviations of the projected values for the expected return, real values, 2) the maximization of the predicted forecasts when the investor expectations are met, with the minimal results when the real return appeared lower than the predicted one. The selection of the learning sample optimal length is performed on the principle of sliding verification of independent material. The described procedure is tested on realistic examples: the US stock market return forecast and staple crop yield gain forecast in the Russian Federation. The suggested algorithm underlying the synthesis of extensively diversified portfolios improves e investment problem solving effectiveness with contribution of considerations of both the markets in focus and the suitable financial instruments. .