PyPortOptimization: A portfolio optimization pipeline leveraging multiple expected return methods, risk models, and post-optimization allocation techniques

Abstract

This paper presents PyPortOptimization, an automated portfolio optimization library that incorporates multiple methods for expected returns, risk return modeling, and portfolio optimization. The library offers a flexible and scalable solution for constructing optimized portfolios by supporting various risk-return matrices, covariance and correlation matrices, and optimization methods. Users can customize the pipeline at every step, from data acquisition to post-processing of portfolio weights, using their own methods or selecting from predefined options. Built-in Monte Carlo simulations help assess portfolio robustness, while performance metrics such as return, risk, and Sharpe ratio are calculated to evaluate optimization results.

• •
  The study compares various configured methods for each step of the portfolio optimization pipeline, including expected returns, risk-modeling and optimization techniques.
• •
  Custom Designed Allocator outperformed. For example, the Proportional Allocator's sharpe ratio of out-performed the expected average.
• •
  A caching system was implemented to optimize execution time.