Scholarlink

<p>The effectiveness of portfolio management can be improved by integrating portfolio selection with stock return prediction. This paper proposes a novel predictive framework, E-BO-XGBoost, leveraging stacked a ensemble model to forecast multi-period stock returns. A predictive multi-period mean semi-absolute deviation credibilistic portfolio model is proposed, which incorporates transaction costs, upper and lower bounds, borrowing constraints, chance constraints, and cardinality constraints. The optimization problem corresponding to the model is a mixed-integer dynamic optimization problem with path dependence, which can be solved using the Genetic Algorithm. To validate the effectiveness of the proposed model, empirical analysis is conducted using the S&amp;P 500 Index constituent stocks, and paired <i>t</i>-tests are applied to compare its returns with those of traditional benchmarks. The results show that the proposed model delivers significantly superior investment performance, and robustness tests based on the CSI 100 Index constituent stocks further corroborate these findings.</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

A Predictive Multi-Period Credibilistic Portfolio Selection Based on Stacked Ensemble

  • 期刊论文

  • 发表日期：

    2025年11月18日

查看全文
  • Peng Zhang,
  • Shiyin Mo,
  • Zibin Peng,
  • Yongquan Zeng

摘要

The effectiveness of portfolio management can be improved by integrating portfolio selection with stock return prediction. This paper proposes a novel predictive framework, E-BO-XGBoost, leveraging stacked a ensemble model to forecast multi-period stock returns. A predictive multi-period mean semi-absolute deviation credibilistic portfolio model is proposed, which incorporates transaction costs, upper and lower bounds, borrowing constraints, chance constraints, and cardinality constraints. The optimization problem corresponding to the model is a mixed-integer dynamic optimization problem with path dependence, which can be solved using the Genetic Algorithm. To validate the effectiveness of the proposed model, empirical analysis is conducted using the S&P 500 Index constituent stocks, and paired t-tests are applied to compare its returns with those of traditional benchmarks. The results show that the proposed model delivers significantly superior investment performance, and robustness tests based on the CSI 100 Index constituent stocks further corroborate these findings.