Intelligent ESG portfolio optimization: A multi-objective AI-driven framework for sustainable investments in the Indian stock market
Environmental, Social, and Governance (ESG) portfolio management impacts the investor in Sustainable and Socially Responsible Investments (SRIs). However, existing frameworks often fail to integrate artificial intelligence (AI) to optimize ESG portfolios, particularly in the Indian stock market. The...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-06-01
|
| Series: | Sustainable Futures |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2666188825003971 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Environmental, Social, and Governance (ESG) portfolio management impacts the investor in Sustainable and Socially Responsible Investments (SRIs). However, existing frameworks often fail to integrate artificial intelligence (AI) to optimize ESG portfolios, particularly in the Indian stock market. They are lacking in terms of achieving higher efficiency in predictive returns and less efficient in integrating ESG scores with portfolio optimization and subsequently rebalancing the optimum portfolio returns. Hence, this study proposes a three-stage AI-driven framework to effectively manage ESG portfolios while maximizing predictive accuracy and investment efficiency. It demonstrates on 30 randomly selected ESG-ranked Indian stocks from diverse sets, and simulates a retail portfolio scenario by leveraging advanced machine learning and optimization techniques. In first stage, a Multivariate Bidirectional Long Short-Term Memory (MBi-LSTM) network is utilized to enhance return prediction accuracy, capturing the market’s nonlinear dynamics. In the second stage, portfolio allocation is optimized by balancing risk, return, and ESG scores using the Non-dominated Sorting Genetic (NSG) algorithm. In the final stage, dynamic portfolio rebalancing is performed using the Actor-Critic Reinforcement Learning (RL) algorithm by integrating ESG scores. For predictive modeling, MBi-LSTM showed the lowest errors in all the parameters. The NSG-enabled portfolio allocation method achieved 68.58 % higher similarity to actual returns than the Mean-Variance Markowitz (MVM) model while maximizing the ESG-Sortino Ratio. In portfolio rebalancing, it outperformed traditional agent-based methods, with returns higher by 120 %, 235 %, and 101 % in average, maximum, and minimum returns, respectively. The effectiveness of the demonstrated Multi-Objective AI-driven framework is showing in every stage validates against the conventional evaluation metrics, showcasing its superior performance and practical viability for sustainable and SRI investment in the Indian stock market. Hence, this framework can help the investor to manage their ESG portfolio for a better SRI. |
|---|---|
| ISSN: | 2666-1888 |