Portfolio Optimization Based on Dynamic Multifactor Stock Scoring

Current portfolio researches focus on the optimization model design and the risk measurement. Specifically, existing methods mainly utilize the return time series itself as an input, while factors describing various aspects of a stock’s state are not fully explored. Moreover, most current multifactor scoring models are linear, and may not adapt the complex market changes in a timely manner. Then, based on the above considerations, this paper presents a comprehensive investment model that seamlessly integrates stock selection and portfolio management into a cohesive framework. By incorporating fundamental and momentum factors, along with the predicted factor derived from technical factors, a nonlinear dynamic stock scoring method is proposed. These scores are then combined together to construct a portfolio optimization model. In this way, the portfolio returns are significantly improved, and the proposed advantages are experimentally verified compared to some state-of-the-art works.