Deep Statistical Fusion of LSTM and ARIMA for ESG-Based Financial Risk and Volatility Forecasting
Dr Nagesh C, Geethanjali K, Mohammed Ghouse S, Ganesh G, Isma Meharaz S
In the era of sustainable finance, Environmental, Social, and Governance (ESG) factors have emerged
as key indicators influencing market behavior and investor sentiment. This study presents a novel hybrid
framework that integrates Long Short-Term Memory (LSTM) networks and Auto-Regressive Integrated
Moving Average (ARIMA) models to forecast financial market volatility and risk while incorporating
ESG signals. The proposed deep statistical fusion model leverages the strengths of ARIMA in capturing
linear temporal dependencies and LSTM’s ability to model complex nonlinear patterns from sequential
data. ESG scores, along with historical price movements and macroeconomic indicators, are used as
primary inputs to enhance model sensitivity to sustainability-related risk. Experiments were conducted
using real-world datasets from global stock indices (e.g., NSE, S&P 500) and third-party ESG rating
providers. The model's performance was assessed using Root Mean Square Error (RMSE), Mean
Absolute Percentage Error (MAPE), and volatility clustering evaluation. The hybrid LSTM-ARIMA
model achieved an RMSE of 1.92 and MAPE of 3.85%, outperforming standalone ARIMA (RMSE:
3.14, MAPE: 6.42%) and LSTM (RMSE: 2.41, MAPE: 5.12%). Additionally, the proposed model
demonstrated better risk sensitivity by accurately flagging high-volatility periods linked to ESG
controversies and macroeconomic disruptions. These results confirm that incorporating ESG factors
within a deep statistical fusion framework enhances forecasting precision, offering a robust tool for
financial institutions and ESG-conscious investors in proactive risk management and strategic decisionmaking.
