Methods of investment portfolio optimization

Abstract

Subject. Evolution and Current State of Investment Portfolio Optimization Methods. Objectives. To develop an investment portfolio optimization model by expanding the system of constraints, which allows taking into account asset liquidity, diversification requirements, and the uncertainty of forecasted return estimates. Methods. Comparative and systems analysis were applied, along with risk theory methods and mathematical modelling techniques. Results. A comparative assessment of key investment portfolio optimization methods — from H. Markowitz’s theory to machine learning algorithms — was carried out; their advantages and limitations were identified. The necessity of modifying existing models to account for risk asymmetry, liquidity, and institutional factors was substantiated. An improved mathematical model is proposed, in which the objective function for maximizing returns is supplemented with constraints on the maximum and minimum shares of assets (to ensure diversification), on the share of asset classes, on liquidity, and which also accounts for forecast uncertainty (illustrated by an LSTM model) through allowable deviations. Conclusions and Relevance. The proposed model enhancements increase its practical applicability and the validity of investment decisions by more fully accounting for contemporary risk and uncertainty factors. The results can be used by financial managers, investment analysts, and risk managers in companies to improve the efficiency of securities portfolio management under volatile market conditions.