Scholarlink

<p>This research introduces a novel ESG-Constrained Multi-Objective Portfolio Optimization Model that integrates higher moments and asymmetric downside risk measures. A two-stage data-driven framework is proposed for asset selection and return prediction. In the first stage, Robust Expectation Maximization (REM) clustering is combined with the TODIM (an acronym in Portuguese for Interactive and Multi-criteria Decision Making) Multi-Criteria Decision-Making (MCDM) method to identify high-potential assets. Unlike conventional approaches that rely solely on clustering, this integration enhances asset selection by incorporating decision-making preferences and risk considerations. Historical financial data from Nifty 100 index companies is used for analysis. In the second stage, a deep learning ensemble model, BN-XGBoost (BiLSTM and N-BEATS ensembled with XGBoost), is employed to predict the future daily returns of the selected assets. The ensemble approach effectively enhances prediction accuracy, as demonstrated by the close alignment between actual and predicted returns. The multi-objective portfolio optimization problem is then solved using the Teaching Learning-Based Optimization (TLBO) algorithm, a metaheuristic approach suited for complex financial landscapes. To assess the impact of ESG (Environmental, Social, and Governance) constraint, two portfolio cases - one with ESG considerations and one without are analyzed. The results highlight the importance of ESG factors in promoting sustainable investment strategies. Finally, to demonstrate the robustness of the proposed model, a comparison with the traditional Mean-Variance (MV) Portfolio Optimization Model is conducted. The findings confirm that the proposed approach achieves superior diversification, risk management, and sustainability alignment, paving the way for further exploration of hybrid clustering, MCDM, and deep learning-based portfolio selection methodologies.</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

Integrated Robust EM and TODIM Approach with Teaching Learning Based Portfolio Optimization

  • 期刊论文

  • 发表日期：

    2025年6月12日

查看全文
  • Veena Jain,
  • Rishi Rajan Sahay,
  • Nupur

摘要

This research introduces a novel ESG-Constrained Multi-Objective Portfolio Optimization Model that integrates higher moments and asymmetric downside risk measures. A two-stage data-driven framework is proposed for asset selection and return prediction. In the first stage, Robust Expectation Maximization (REM) clustering is combined with the TODIM (an acronym in Portuguese for Interactive and Multi-criteria Decision Making) Multi-Criteria Decision-Making (MCDM) method to identify high-potential assets. Unlike conventional approaches that rely solely on clustering, this integration enhances asset selection by incorporating decision-making preferences and risk considerations. Historical financial data from Nifty 100 index companies is used for analysis. In the second stage, a deep learning ensemble model, BN-XGBoost (BiLSTM and N-BEATS ensembled with XGBoost), is employed to predict the future daily returns of the selected assets. The ensemble approach effectively enhances prediction accuracy, as demonstrated by the close alignment between actual and predicted returns. The multi-objective portfolio optimization problem is then solved using the Teaching Learning-Based Optimization (TLBO) algorithm, a metaheuristic approach suited for complex financial landscapes. To assess the impact of ESG (Environmental, Social, and Governance) constraint, two portfolio cases - one with ESG considerations and one without are analyzed. The results highlight the importance of ESG factors in promoting sustainable investment strategies. Finally, to demonstrate the robustness of the proposed model, a comparison with the traditional Mean-Variance (MV) Portfolio Optimization Model is conducted. The findings confirm that the proposed approach achieves superior diversification, risk management, and sustainability alignment, paving the way for further exploration of hybrid clustering, MCDM, and deep learning-based portfolio selection methodologies.