Paper Title
An ESG-Aware Hybrid CNN–LSTM–Attention Framework With Uncertainty-Driven Portfolio Optimization on Bombay Stock Exchange Equities

Abstract
Sustainable investing has emerged as a critical paradigm in modern financial markets, integrating environmen-tal, social, and governance (ESG) considerations with traditional financial analysis. This paper proposes a comprehensive ESG-aware deep learning framework for stock return prediction and portfolio optimization using equities listed on the Bombay Stock Exchange (BSE). A hybrid Convolutional Neural Network–Long Short-Term Memory with Attention (CNN–LSTM–A) architec-ture is employed to capture both short-term market microstruc-tures and long-term temporal dependencies. The framework in-tegrates multi-task learning for joint price and return prediction, Monte-Carlo Dropout for predictive uncertainty estimation, and an uncertainty-aware portfolio optimization strategy. Empirical results demonstrate improved prediction stability, reduced draw-downs, and superior risk-adjusted portfolio performance, vali-dating the effectiveness of the proposed approach for sustainable investing in emerging markets. Keywords - ESG Investing, Deep Learning, CNN-LSTM, Attention Mechanism, Monte-Carlo Dropout, Portfolio Optimiza-tion, BSE