A machine learning project to predict stock price movements using historical data and financial indicators. Built with Python, Spark, and modern deep learning techniques.
Accurately forecasting stock prices can provide valuable insights for investors and traders. However, this is a challenging task due to nonlinear dynamics in the market. This project applies advanced algorithms to uncover patterns in the complex interactions between shares and macroeconomic factors.
A Streamlit web app is provided to demo model performance in real-time
Daily stock data for TSLA from 2010-present obtained from Yahoo Finance. Macroeconomic indicators like S&P 500 from FRED are also included. Over 10 years of historical market context is analyzed.
- Linear Regression
- Random Forest
- Gradient Boosted Trees
- Long Short-Term Memory Network (LSTM)
The LSTM Neural Network achieved the best performance with a RMSE of 0.057 on the test set, outperforming other algorithms.
Instructions provided to run the full codebase locally or on Databricks for reproducibility. The web app allows interacting with a live model.
Enhancements like adding new deep learning architectures and risk analysis are planned for future iterations. Contributions welcome!