📊 Dynamic Portfolio Allocation under Uncertainty

This project implements and compares dynamic portfolio optimization methods for allocating wealth between risky and a risk-free asset over multiple periods. It explores both analytical and approximate solutions under different investor utility functions, including log and CRRA utilities.

---

🔍 Project Overview

• Objective: Determine optimal portfolio weights over a finite investment horizon in the presence of uncertainty.
• Frameworks Used:
  • Monte Carlo simulation
  • Convex optimization via CVXPY
  • Approximate Dynamic Programming (ADP)
• Key Features:
  • Exact dynamic programming with log utility and multi-asset convex optimization
  • Approximate dynamic programming with a single risky asset, CRRA utility, and cubic interpolation
  • Wealth grid construction and policy extraction

---

📌 Modeling Assumptions

• The market is assumed to be frictionless, with no transaction costs, slippage, or liquidity constraints.
• Risk-free and risky assets are perfectly divisible.

---

📁 Contents

• dynamicPortfolioAllocation.py: Core module with reusable functions for:
  • Return simulation
  • Optimal portfolio computation
  • Value function approximation
• dynamic_portfolio_allocation.ipynb: Jupyter notebook demonstrating:
  • Examples of portfolio allocation under various scenarios
  • Visualizations of value functions and optimal policies
  • Comparison between exact and approximate solutions

---

🛠 Tools & Technologies

---

📈 Key Results

• Demonstrates how optimal asset allocations change over time and across wealth levels.
• Highlights the performance of approximate dynamic programming versus exact methods.
• Offers a blueprint for extending the ADP to multi-asset, multi-period portfolio problems.

---

📚 Citation

If you use this code in your work, please cite this repository:

https://github.com/ramiuness/dynamic-portfolio-allocation

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.