Scholarlink

<p>This study introduces a dynamic factor mining framework to address limitations in traditional approaches, such as model degradation under market non-stationarity and challenges in multi-objective optimization. By integrating Genetic Programming with a tri-objective fitness function maximizing Sharpe ratio, annualized return, and minimizing maximum drawdown—the model evolves interpretable alpha factors within a rolling window mechanism. These factors are further processed using LightGBM for nonlinear ensemble learning and optimized via mean-variance allocation. Empirical validation on the CSI 300 Index (March 2020–June 2024) shows superior performance: cumulative return of 416.38%, annualized return of 47.75%, Sharpe ratio of 1.59, and maximum drawdown below 26%. The results demonstrate the framework’s effectiveness in adapting to evolving market conditions and balancing competing financial objectives.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Dynamic Factor Mining Based on Multi-Objective Fitness and Its Empirical Study in Multi-Factor Strategies

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • Wang Yuxue,
  • Geng Shuanghong

摘要

This study introduces a dynamic factor mining framework to address limitations in traditional approaches, such as model degradation under market non-stationarity and challenges in multi-objective optimization. By integrating Genetic Programming with a tri-objective fitness function maximizing Sharpe ratio, annualized return, and minimizing maximum drawdown—the model evolves interpretable alpha factors within a rolling window mechanism. These factors are further processed using LightGBM for nonlinear ensemble learning and optimized via mean-variance allocation. Empirical validation on the CSI 300 Index (March 2020–June 2024) shows superior performance: cumulative return of 416.38%, annualized return of 47.75%, Sharpe ratio of 1.59, and maximum drawdown below 26%. The results demonstrate the framework’s effectiveness in adapting to evolving market conditions and balancing competing financial objectives.