Multi-Period Investment Strategy Based On Receding Horizon Optimization And Black-Litterman Model

Portfolio optimization is a critical and extensively studied area in finance that aims to maximize returns while minimizing risks using statistical,modeling,and optimization techniques.Through the applications of these techniques,the dynamic asset allocation problem in the investment process is transformed into a typical optimization problem.Real-world investment problems are influenced by various factors,such as dynamic fluctuations in asset values and transaction processes which are subject to various constraints and costs.The multi-period portfolio problem,which takes into account the above factors comprehensively over future investment periods,is more aligned with the dynamic and real-time nature of the investment market.Compared to the single-period portfolio optimization problem,the multi-period portfolio optimization problem yields superior investment strategies.Additionally,the Black-Litterman(BL)model is well-suited to incorporate investor’s personal views into the investment process,forming a strategy that better aligns with the user’s investment preferences.Therefore,the multiperiod portfolio optimization problem based on the BL model has received widespread attention.This dissertation studies the multi-period portfolio optimization problem based on the BL model.By incorporating asset returns,risks,transaction costs and other constraints into the problem description,a comprehensive optimization problem is formed.The receding horizon optimization method is then used to solve the problem.At each time step,the first solution in the future optimization solution sequence is implemented,and the subsequent solutions are discarded.When the next time step arrives,the process is repeated to form a new optimization problem and find the future optimization solution sequence,and the first solution is implemented again.This approach can provide real-time dynamic optimization allocation strategies which are more in line with the real financial market,as dynamic market information and investor views can be incorporated into the optimization problem in real-time at each time step.The main work of this dissertation includes:(1)To address the issue of single-period investment portfolios not accounting for complex investment scenarios,this dissertation proposes a solution in the form of a multi-period investment asset allocation strategy.By dividing the investment process into multiple stages and considering factors such as costs and constraints,a multi-period investment portfolio problem is designed with the goal of maximizing returns and minimizing risks.Risk estimates are derived through the Fama-French 5 factor model,and a receding horizon method is applied to obtain the optimal investment portfolio strategy.The multi-period horizon is chosen to be a shorter period rather than the entire trading range to maintain the controllability and dynamism of the optimization problem.(2)To address the issue of the BL model requiring investors to manually provide view confidence levels,which makes it difficult to incorporate into multi-period optimization process,a data-driven approach is applied to achieve dynamic view confidence levels.The approach utilizes a regime-switching model to predict two market states and employs a sigmoid activation function to derive dynamic view confidence levels from the predicted states.Experimental results demonstrate that this method can effectively provide dynamic confidence levels for investment views provided.(3)To overcome the limitation of multi-period optimization models only considering market factors while neglecting investor subjectivity,returns and risks are estimated based on the BL model and incorporated into a multi-period optimization process.The optimal trading strategy is then obtained using a receding horizon optimization method.This approach,which effectively captures investors’ subjective intentions,is of significant importance in practical applications as it provides a better balance between market and individual investor perspectives.