Optimal portfolio for nonstationary security market

A new interpretation of the classical portfolio problem is suggested. It is based on the optimization on the conditional expectation of a utility function where the conditional expectation is calculated taking into account all information available at the moment of decision making. This approach allows us to show that the risk is a sequence of forecasting errors. Main classes of forecasting algorithms are described and compared. A similarity between some variants of solutions to the modified portfolio problem and the traditional technical analysis recommendations is shown. The scheme is essentially important for nonstationary nonequilibrium security market. Some results of its practical application to the Russian bond market are presented.