Factor Model Explanation

1. Overview: This project contains a factor model implementation designed for financial data analysis, specifically for constructing and analyzing portfolios based on various financial factors.

2. Files and Classes:

• main.py

  • This is the entry point of the application. It initializes the portfolios, performs data preparation and cleaning, standardizes features, executes a grid search to find optimal parameters for Lasso and ElasticNet models, and finally performs regression analyses using Linear, Lasso, Ridge, and ElasticNet models.

• portfolio.py

  • An abstract base class that defines the structure for portfolio classes. It requires the implementation of initialize_components() and aggregate_returns() methods.

• contract_portfolio.py

  • A subclass of Portfolio that is specialized for managing a portfolio of contracts. It implements methods to initialize contract components and aggregate their returns.

• factor_portfolio.py

  • Another subclass of Portfolio, tailored for factor portfolios. It initializes factor components and aggregates their returns, similar to ContractPortfolio.

• factor.py

  • Defines the Factor class, which is responsible for fetching and processing financial data for a given ticker. It includes methods to download historical data, calculate returns, and access the processed data.

• portfolio_utils.py

  • Contains utility functions for data preparation and regression analysis:

    • prepare_data(): Joins and cleans portfolio and factor returns data.
    • perform_grid_search(): Finds the optimal alpha parameters for Lasso and ElasticNet.
    • perform_different_regressions(): Fits Linear, Lasso, Ridge, and ElasticNet models.
    • plot_regression_results(): Plots the regression results on a given axes object.
    • calculate_adjusted_r_squared(): Calculates the R-squared and adjusted R-squared values.

3. Model Workflow

• The main.py script initiates the process by creating portfolios and retrieving their returns.
• The data is then cleaned and standardized.
• Optimal hyperparameters for certain regression models are determined via grid search.
• Different types of regressions are performed.
• The results, including regression plots, are displayed for analysis.