With over 2800 companies listed as stocks in the United States alone, it’s hard for someone without the current knowledge to know which stocks are the best options to buy. This project, started by Professor Liu, will determine which of six different regression models is the best in predicting the next best stock to buy.
Tools to be Utilized
- Visual Studio Code
- Python
- Pandas
- NumPy
- Matplotlib
- Swig
- box2d
- Ski-kit Learn
Semester Goals
- Acquire Six Functional Regression Models Across Various Finance Datasets:
- Obtain six operational regression models suitable for finance datasets.
- Models include linear, polynomial, ridge, lasso, logistic, and Bayesian regression.
- Prepare and Structure Datasets for Model Training and Testing:
- Ensure datasets from diverse sources like Yahoo Finance, and the FinRL Competition (2023) are cleansed and organized.
- Enable seamless training and testing procedures for the models.
- Initiate Model Training:
- Train the initial set of regression models using the prepared datasets.
- Ensure models are adequately primed for subsequent evaluation.
- Implement Backtesting and Error Evaluation for Initial Models:
- Construct a robust backtesting framework to assess model performance.
- Identify and evaluate errors within the initial model predictions.
- Apply Cross-Validation Techniques to Mitigate Overfitting in Backtesting:
- Implement cross-validation methodologies to prevent overfitting during backtesting.
- Ensure the reliability and accuracy of model evaluations.
- Enhance Models with Additional Predictors:
- Incorporate supplementary predictors to refine model accuracy and predictive capabilities.
- Explore ensemble learning techniques to amalgamate insights from multiple predictors effectively.