A hybrid approach to stock price prediction using Normal Inverse Gaussian (NIG) and Geometric Brownian Motion (GBM) models combined with LSTM neural networks.
This repository explores the use of stochastic processes and deep learning for stock price prediction, implementing models that integrate mathematical finance and modern machine learning.
Predicting stock prices is challenging due to market volatility and the influence of external factors like sentiment and macroeconomic trends. This project uses a combination of:
- Normal Inverse Gaussian (NIG): A probabilistic model to capture asymmetric volatility.
- Geometric Brownian Motion (GBM): A classic stochastic model for simulating price paths.
- Long Short-Term Memory (LSTM): A deep learning model designed for sequential data like stock prices.
This project features advanced simulation models such as the Normal Inverse Gaussian (NIG), which captures price dynamics with non-Gaussian features, and the Geometric Brownian Motion (GBM), which simulates stock price paths based on historical data. It also incorporates deep learning predictions using LSTM, which excels at capturing long-term dependencies in sequential stock data. Evaluation metrics like Mean Absolute Percentage Error (MAPE) and Mean Squared Error (MSE) are used to assess model performance. Additionally, the project generates comparative plots to visualize the alignment of simulated and actual stock price paths.
Contributions are welcome! Feel free to submit a pull request with model improvements.