Stock-market-forecasting presents a trading strategy utilizing LSTM and random forest networks to predict intraday stock price movements. The project aims to analyze the effectiveness of these models in forecasting directional movements of S&P 500 stocks for intraday trading, drawing upon historical data from 1993 to 2018. The core problem addressed is to develop a profitable trading strategy based on sophisticated machine learning techniques.
This project demonstrates a practical application of LSTM and Random Forest models for financial forecasting, validated on a comprehensive historical dataset. The inclusion of detailed plots analyzing model performance metrics and feature importance aids in understanding model behavior. The results show that the LSTM and Random Forest models outperform single-feature analysis.
- Model Implementation: Implements LSTM and Random Forest models for directional movement prediction.
- Data Analysis: Includes plots of cumulative money growth, average daily returns, and Sharpe ratio.
- Hyperparameter Tuning: Explores and documents the hyperparameter tuning process for the LSTM model.
- Feature Importance: Identifies and analyzes the importance of individual features in the trading strategy.
- Reproducibility: Provides code and instructions for replicating the experiments described in the paper.
- Dependencies: Specifies required Python libraries like scikit-learn and tensorflow.
- Research Paper Reference: Links to the published research paper detailing the methodology and results.
The project is based on published research and includes well-documented code and results. The code is functional and provides insights into the model's performance. The project has a clear research direction and demonstrates a good understanding of the underlying concepts, considering various LSTM architectures and comparing them to simpler GRU models.
This project is valuable for researchers and practitioners interested in applying machine learning to stock market forecasting. It offers a practical implementation of LSTM and Random Forest models, along with detailed analysis and results. The work provides a framework for developing data-driven trading strategies and offers insights into the effectiveness of different modeling approaches compared to manual analysis or simpler models.