High-Dimensional Portfolio Optimization Based on Tree-Structured Factor Model

Certain characteristics are known for their predictability of cross-sectional expected stock returns. We propose a novel portfolio optimization procedure to incorporate multiple characteristic information, which requires minimum parameters and no stringent assumptions. Instead of investing in individual stocks, we conduct portfolio optimization on a large panel of characteristic portfolios generated by a tree-structured portfolio sorting method, which can capture the non-linearities and interactions between stocks' characteristics to help predict the expected returns. Simulations demonstrate that our tree-structured Lasso-based mean-variance (MV) strategy has better out-of-sample Sharpe ratios than the three benchmark strategies: the stock-based MV strategy, the double-sorted portfolio-based MV strategy, and the tree-structured portfolio-based MV strategy.We use daily stock data from the Chinese A-share market from 2001 to 2021 to compare the out-of-sample performances of the tree-structured Lasso-MV with other classical strategies, such as global minimum variance (GMV) and equally weighted (EW) strategies. The empirical results suggest that the tree-structured Lasso-MV strategy can achieve both a higher Sharpe ratio and a smaller standard deviation. These results prove to be robust under different parameter combinations