ML-Enhanced Portfolio Risk Analyzer

Table of Contents

1. Overview
2. Features
3. Setup and Installation
4. Usage Instructions
5. Important Notes
6. Technical Details
7. Future Enhancements

---

Overview

The ML-Enhanced Portfolio Risk Analyzer is a robust tool designed for portfolio risk assessment, optimization, and financial insights. It leverages machine learning models, advanced financial metrics, and interactive data visualizations to help users make informed investment decisions.

---

Features

1. Portfolio Risk Analyzer

• Analyze risk for a custom-selected portfolio.
• Metrics include annualized volatility, Sharpe ratio, maximum drawdown, and more.
• Generates visualizations such as risk gauges, correlation heatmaps, and drawdown graphs.

2. AI Portfolio Assistant

• Chat interface to answer portfolio and investment-related queries.
• Powered by OpenAI GPT-4 for comprehensive financial advice.

3. Voice AI Assistant

• Voice-activated AI for portfolio guidance.
• Records, transcribes, and responds to user queries with audio playback.

4. Portfolio Model Admin Dashboard

• Allows admins to update datasets, retrain models, and manage portfolio analysis pipelines.
• Fetches and integrates stock data and economic indicators in real-time.

5. Portfolio Optimization

• Suggests portfolios based on user-defined risk preferences.
• Incorporates predicted volatility using advanced machine learning models.

---

Setup and Installation

1. Clone the Repository

git clone https://github.com/senlerk/ML-Enhanced-Portfolio-Risk-Analyzer
cd ML-Enhanced-Portfolio-Risk-Analyzer

2. Install Dependencies

pip install -r requirements.txt

3. Set Up Secrets

Create a .streamlit/secrets.toml file in the project directory with the following content:

[admin]
password = "your_admin_password"

[api_keys]
openai_api_key = "your_openai_api_key"
fred_api_key = "your_fred_api_key"

4. Run the Application

streamlit run app.py

---

Usage Instructions

Initial Setup

1. Error on First Run: If you encounter the following error on the first run:

   Error loading ML components: [Errno 2] No such file or directory: 'random_forest_model.pkl'
   Failed to load ML components. Some features may be unavailable.
   

   This happens because the application doesn't have a pre-trained model loaded initially.

2. Resolve the Error:

   • Navigate to the Admin Dashboard page after starting the application.
   • Enter the Admin Password set in your .streamlit/secrets.toml file.
   • On the Model Training tab, run the steps to gather data, merge data, and train the model.
   • The application will save the required model files (random_forest_model.pkl, scaler.pkl, etc.).

3. Restart Application:

   • Restart the application after training the model to ensure all features are fully functional.

---

Important Notes

• Admin Access Required:

  • The Admin Dashboard requires a password to ensure secure access to critical features like model training.
  • Use the password defined in the .streamlit/secrets.toml file.

• API Keys:

  • Ensure valid API keys for OpenAI and FRED are added to the .streamlit/secrets.toml file. Without them, certain functionalities will be unavailable.

---

Technical Details

Tech Stack

• Frontend:
  Streamlit is used for creating an interactive and visually appealing user interface. The platform simplifies deployment and ensures cross-platform compatibility.

• Backend:
  Python-based backend using:

  • APIs: Integration with OpenAI API (GPT-4) for AI-powered assistant functionalities and FRED API for economic indicators.
  • Data Libraries:
    • yfinance: Fetch real-time stock data for S&P 500 companies.
    • fredapi: Access Federal Reserve data.
    • BeautifulSoup: Scrape additional stock information if needed.

• Machine Learning:
  Random Forest Regressor model for predicting portfolio volatility. Includes advanced feature engineering with lagging, rolling averages, and interaction terms. ML pipeline uses scikit-learn for preprocessing, training, and model evaluation.

• Visualization:
  Interactive charts powered by Plotly, including:

  • Risk Gauges
  • Correlation Heatmaps
  • Portfolio Composition Pie Charts
  • Time Series Analysis

• Optimization:
  Portfolio optimization using scipy.optimize.minimize to calculate efficient allocations based on user-defined risk thresholds.

• Data Sources:

  • S&P 500 Historical Stock Data: Pulled from yfinance.
  • FRED Economic Indicators: Includes VIX, Consumer Sentiment Index, Yield Curve Spreads, and more.

• Deployment:
  Dockerized application deployed on Google Cloud Run with the following configurations:

  • Platform: Managed
  • Region: us-central1
  • Resources: Increased RAM to 4GB for heavy computational tasks.
  • Authentication: Application allows public access with Google Cloud Run settings.

---

Future Enhancements

1. Integration of More ML Models:

   • Explore additional algorithms like Gradient Boosting or Deep Learning for enhanced predictive accuracy.

2. Advanced Portfolio Optimization:

   • Implement methods like Monte Carlo Simulations or Black-Litterman for improved allocation strategies.

3. Real-Time Data Integration:

   • Fetch real-time financial data for instant analysis and insights.

4. User Management System:

   • Implement user accounts and role-based access for multi-user scenarios.

5. Performance Enhancements:

   • Utilize faster data handling libraries like Dask for large datasets.