A Novel Hybrid Monte Carlo-Random Forest Framework for Improved Financial Predictions

Abstract

In an era of increasing market volatility, accurate predictive tools are critical to navigate financial uncertainties. This study introduces a novel hybrid ensemble model that integrates Random Forest Regressors (RFRs) with Monte Carlo (MC) simulations to enhance predictive accuracy and quantify uncertainties. The hybrid model combines RFRs’ ability to capture complex patterns with MC simulations’ strength in modeling random price movements and volatility patterns. Specifically, the model generates multiple price paths using MC simulations based on historical volatility and drift, while the RFR component provides robust predictions by aggregating decision trees trained on historical price data. The integration is achieved by combining the RFR predictions with the ensemble average of MC-simulated price paths, creating a hybrid output that balances deterministic and probabilistic insights. The model effectively captures volatility clustering by reflecting historical volatility patterns in simulated paths and models stochastic trends through random sampling of future price movements. When applied to Disney stock prices and the S&P 500 index, the model demonstrated significant improvements in predictive performance over traditional methods, with reductions in the Mean Absolute Error (MAE). The results underscore its ability to capture volatility clustering and stochastic trends, providing practical advantages for portfolio optimization and risk assessment. To address the computational trade-off, the model employs parallel processing and optimized simulation parameters, ensuring scalability for real-world applications. Although computationally intensive, the hybrid approach presents a reliable framework, paving the way for advances in financial forecasting methodologies.