azapy project

Financial Portfolio Optimization Algorithms

An open-source python library for everybody

Author: Mircea Marinescu

email: Mircea.Marinescu@outlook.com

Package documentation

Package installation: pip install azapy

A graphical user interface is provided by azapyGUI package.

Contents

A. Risk-based portfolio optimization algorithms

1. mCVaR - mixture CVaR (Conditional Value at Risk)
2. mSMCR - mixture SMCR (Second Moment Coherent Risk)
3. mMAD - m-level MAD (Mean Absolute Deviation)
4. mLSD - m-level LSD (Lower Semi-Deviation)
5. mBTAD - mixture BTAD (Below Threshold Absolute Deviation)
6. mBTSD - mixture BTSD (Below Threshold Semi-Deviation)
7. GINI - Gini index (as in Corrado Gini statistician 1884-1965)
8. SD - standard deviation
9. MV - variance (as in mean-variance model)
10. mEVaR - mixture EVaR (Entropic Value at Risk) <span style="color:blue">(beta version)</span>

For each class of portfolio the following optimization strategies are available

1. Optimal-risk portfolio for targeted expected rate of return value
2. Sharpe-optimal portfolio - maximization of generalized Sharpe ratio
3. Sharpe-optimal portfolio - minimization of inverse generalized Sharpe ratio
4. Minimum risk portfolio
5. Optimal-risk portfolio for a fixed risk-aversion factor
6. Optimal-risk portfolio with the same risk value as a benchmark portfolio (e.g., same as equal weighted portfolio)
7. Optimal-diversified portfolio for targeted expected rate of return (minimization of inverse 1-D ratio) <span style="color:blue">(beta version)</span>
8. Optimal-diversified portfolio for targeted expected rate of return (maximization of 1-D ratio) <span style="color:blue">(beta version)</span>
9. Maximum diversified portfolio <span style="color:blue">(beta version)</span>
10. Optimal-diversified portfolio with the same diversification factor as a benchmark portfolio (e.g., same as equal weighted portfolio) <span style="color:blue">(beta version)</span>
11. Optimal-diversified portfolio with the same expected rate of return as a benchmark portfolio (e.g., same as equal weighted portfolio) <span style="color:blue">(beta version)</span>

B. "Naïve" portfolio strategies

1. Constant weighted portfolio. A particular case is equal weighted portfolio.
2. Inverse volatility portfolio (i.e., portfolio weights are proportional to the inverse of asset volatilities)
3. Inverse variance portfolio (i.e., portfolio weights are proportional to the inverse of asset variances)
4. Inverse drawdown portfolio (i.e., portfolio weights are proportional to the asset absolute value of maximum drawdowns over a predefined historical period)

C. Greedy portfolio optimization strategies

1. Kelly's portfolio (as in John Larry Kelly Jr. scientist 1923-1965) - maximization of portfolio log returns
2. Universal portfolio (Thomas M. Cover 1996) <span style="color:blue">(beta version)</span>

D. Market Selectors

1. Dual Momentum Selector <span style="color:blue">(beta version)</span>
2. Correlation Clustering Selector <span style="color:blue">(beta version)</span>

Utility functions:

• Collect historical market data from various providers.

  Supported providers:

  • yahoo.com
  • eodhistoricaldata.com
  • alphavantage.co
  • marketstack.com

• Generate business calendars for all major exchanges across the world.

• Generate rebalancing portfolio schedules.

• Append a cash-like security to an existing market data object.

• Update market data saved in a directory.

• N-simplex random vectors generators.

Third-party packages used by azapy 1.2.4

• python 3.11.8
• pandas 2.1.4
• numpy 1.26.2
• scipy 1.11.4
• statsmodels 0.14.0
• matplotlib 3.8.0
• plotly 5.9.0
• requests 2.31.0
• pandas_market_calendars 4.3.2
• ecos 2.0.12
• cvxopt 1.3.2
• ta 0.11.0
• yfinance 0.2.33