Using Deep Learning to Build Prediction Models of Carbon Emissions and Leadership Ethics for Stock Returns - Based on BERT and LSTM Models

This study applies text analysis and deep learning to better understand the effect of carbon emissions on stock time-series data to achieve sustainability goals through leadership ethics. The study considers leadership ethics climate change policy by text analysis and uses XGBoost and Long Short-Term Memory (LSTM) to build a stock return predictive model with carbon emission management. A key contribution of this research is the exploration of CEO ethical orientation (deontology versus utilitarianism) in influencing corporate carbon policies, which highlights the duality between maximizing utility and reducing emissions. Additionally, it provides empirical evidence on the role of leadership ethics in sustainable finance, filling an important gap in the literature on the intersection of ethical leadership and financial performance. The results support the better predictive accuracy of XGBoost, especially when considering CEO ethics, offering new insights into how leadership can shape corporate responses to climate risks and stock market outcomes.