State-Dependent Online Return Prediction and Decision Making

This chapter studies how to incorporate the state-dependent mechanism into online return prediction and investment decision making. Transaction cost is the first concern this chapter tries to tackle by deriving the analytical transaction remainder factor (TRF). In the meanwhile, the assets’ market state switching in practical investment activities is considered. Based on this, the exponential moving average approach with state-dependent properties (SEMA) is proposed, capable of precisely forecasting the returns of assets by leveraging their historical return records and current market conditions. Then, according to different market states, the online portfolio optimization model with and without risk parity constraint are respectively studied, and the state-dependent online portfolio selection algorithm (SOPS) is designed to enhance the profitability of strategy. Empirical validation shows that the newly introduced SOPS algorithm is capable of outperforming a wide range of cutting-edge OLPS algorithms.