Portfolio Optimization Tool

Introduction

The Portfolio Optimization Tool is a project aimed at addressing financial portfolio allocation challenges using metaheuristic optimization algorithms. It leverages historical stock price data from the Australian Securities Exchange (ASX) to analyze and optimize portfolios, providing a user-friendly interface for visualization and decision-making. We think that a future roadmap of improvements, would involve making this tool integrate easily with other csv historical data from different datasets and historical market info.

Project Made with:

• João Rebelo
• Gonçalo Sousa

GUI Screenshot

Image 1:GUI Interface

Installation

1. Set up a virtual environment:

   python -m venv venv

2. Activate the virtual environment:

• Windows:

  venv\Scripts\activate

• Unix based OS:

  source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

Implemented Features

Optimization Algorithms

The tool includes six meta-heuristic algorithms, each with customizable parameters that allow tinkering with the algorithms :

• Hill Climbing: A local search algorithm.
• Simulated Annealing: A probabilistic global optimization method.
• Tabu Search: A memory-based optimization technique that avoids local optima.
• Genetic Algorithm: An evolutionary approach for optimization.
• Particle Swarm Optimization: A swarm intelligence-based method.
• Differential Evolution: A population-based optimization technique.

Algorithm	Type	Best Use Case	Key Parameter
Hill Climbing	Local Search	Quick analysis	Step size
Simulated Annealing	Global Optimization	Complex problems	Cooling rate
Tabu Search	Memory-based	Avoiding solution cycling	Tabu tenure
Genetic Algorithm	Evolutionary	Multi-objective problems	Mutation rate
Particle Swarm	Swarm Intelligence	Continuous optimization	Inertia weight
Differential Evol.	Population-based	Noisy environments	Crossover rate

Interactive Visualizations

Gain insights through interactive visualizations, including:

• Portfolio Allocation: Displays asset weights in a pie chart.
• Efficient Frontier: Plots risk-return trade-offs.
• Optimization Progress: Tracks algorithm convergence.
• Return vs. Risk Comparison: Compares asset and portfolio metrics.
• Correlation Matrix: Shows asset correlations.

Additional Features

• Constraint Options: Set constraints like risk tolerance, position limits, and a checkbox that allows short selling of stock (be careful this is risky!).
• Performance Testing: Compare and test algorithm performance with single test runs or parallel tests.
• Export Functionality: Save results as CSV or PDF reports.

Testing Framework and Strategy

• Our portfolio optimization tool employs a robust testing framework that runs each algorithm through multiple iterations (50 for single algorithm tests, 5 per algorithm for comparative tests).
• This iterative approach is essential because metaheuristic algorithms contain stochastic elements that can lead to different solutions on each run.
• By executing multiple iterations, we can evaluate the algorithms not just on their average performance metrics (returns, risk, and Sharpe ratio) but also on their consistency and reliability.
• This methodology provides users with statistical confidence in the results, revealing which algorithms deliver the best risk-adjusted returns with the lowest variability.
• The framework also tracks execution times, offering insights into computational efficiency—a critical factor when optimization needs to be performed frequently or with large datasets.
• Users can visualize these comparative results or export them as CSV/PDF reports for further analysis.

How to Use the Program

1. Loading Data

• Click "Add CSV File" to load individual stock price data files.
• Click "Add Folder of CSV Files" to load all CSV files from a directory.

2. Setting Constraints

• Initial Investment: Set your total investment amount.
• Risk Aversion: Set between 0 (risk-seeking) and 1 (risk-averse).
• Max Asset Weight: Maximum percentage any single asset can have in the portfolio.
• Min Return/Max Risk: Set minimum return and maximum risk constraints.
• Allow Short Selling: Check to allow negative weights (short positions).

3. Algorithm Selection and Parameters

Select an algorithm tab and set its parameters: This parameters are based on each individual algorithm implementation, but the user may also change the constraints and this constraints are a common base for running all the different algorithms that impact the output results.

4. Running the Optimization

• Click "Run Portfolio Optimization" to execute the selected algorithm.
• Click "Test Portfolio Optimization" to run multiple tests of the algorithm.
• Check "Test All Algorithms" to compare all algorithms' performance.

5. Viewing Results

The visualization area shows different views based on selection:

• Portfolio Allocation: Pie chart of asset weights.
• Efficient Frontier: Risk-return plot.
• Optimization Progress: Algorithm convergence.
• Return-Risk Comparison: Asset vs. portfolio metrics.
• Correlation Matrix: Asset correlations.

Results summary shows key metrics and allocation details.

Image 2:Portfolio Allocation

Image 3:Efficient Frontier

6. Saving Results

• Click "Save Results" to export allocation to CSV.
• Click "Export to PDF" to save visualizations and results as PDF.

Data Requirements in order to apply the optimization model/tool

CSV files must contain the following columns:

• Date: YYYY-MM-DD format.
• Open: Daily opening price.
• High: Daily high price.
• Low: Daily low price.
• Close: Daily closing price (required).
• Adj Close: Adjusted closing price (required).
• Volume: Trading volume.

Output Reports

• CSV Export: Save portfolio allocation and summary results to a CSV file.
• PDF Export: Generate a detailed PDF report with visualizations and results.