How helebkat/msml604-final-project works
Project type and purpose
• This is a batch data‐analytics pipeline (Python scripts + Jupyter notebook) for portfolio risk analysis:
• Retrieve historical stock data
• Compute daily returns
• Fit univariate (skew‐t & Student’s t) and multivariate Student‐t distributions
• Calculate VaR/CVaR
• Solve a multi‐objective optimization (Pareto frontier)
• Produce summary statistics and plots
File structure
• Top‐level:
• MSML_604_Group_Project.ipynb – narrative and orchestration (optional)
• README.md, proposal.md – documentation
• source/ – six numbered Python modules (1–6) forming the pipeline
• output/ – CSV, NPZ, and summary files created at each stage
• resources/ – reference papers
• No external services or databases—pure file I/O.
• Each script reads one or more input artifacts and writes outputs consumed by the next script.
Extract architectural insights from the README
• Sequential ordering: 1 → 2 → 3 → 4 → 5 → 6
• Key outputs: stock_data.csv ⇒ daily_returns.csv ⇒ fitted‐distribution parameters & samples ⇒ portfolio risk measures ⇒ Pareto‐frontier plot
1. Data Retrieval Module (1-retrieve-data.py)
2. Returns Computation Module (2-compute-returns.py)
3. Univariate MLE Module (3-fit-mle.py)
4. Multivariate Student‐t Module (4-multivariate-student.py)
5. Risk Calculation Module (5-calculate-var-cvar.py)
6. Optimization/Visualization Module (6-pareto-frontier.py)
7. Notebook / Orchestrator (MSML_604_Group_Project.ipynb) – optional high‐level driver
• Raw stock data CSV (stock_data.csv)
• Daily returns CSV (daily_returns.csv)
• Distribution parameters & PDF samples (in‐memory & as npz if saved)
• VaR & CVaR values (console output or CSV)
• Pareto frontier plot
1. Module 1 → stock_data.csv
2. stock_data.csv → Module 2 → daily_returns.csv
3. daily_returns.csv → Module 3 → Univariate fit parameters & samples
4. daily_returns.csv + univariate parameters → Module 4 → Multivariate params (mean, Σ, ν, C)
5. multivariate params + example weights → Module 5 → VaR, CVaR
6. daily_returns.csv + multivariate params → Module 6 → Pareto frontier plot
d. Architectural patterns / principles:
• Linear pipeline / ETL style
• Modular scripts with single responsibility
• Data‐driven (files as contracts between modules)
• Reproducibility: numbered steps
e. Technologies / frameworks:
• Python ≥3.x, pandas for CSV I/O, numpy & scipy for stats, matplotlib for plots
• Jupyter Notebook for orchestration / narrative