📊 Portfolio Optimization using LSTM & Modern Portfolio Theory

🚀 Overview

This project combines Deep Learning (LSTM) with Modern Portfolio Theory (MPT) to build an intelligent portfolio optimization system. It predicts future stock returns using time-series forecasting and optimizes asset allocation to maximize returns while analyzing risk.

🎯 Objectives

• Predict stock returns using LSTM (Deep Learning)
• Apply Modern Portfolio Theory (MPT) for optimization
• Maximize Sharpe Ratio (risk-adjusted return)
• Analyze risk vs return trade-offs

🧠 Tech Stack

• Python
• NumPy, Pandas
• Matplotlib, Seaborn
• yFinance
• TensorFlow / Keras (LSTM)
• SciPy (SLSQP Optimization)

📂 Dataset

• Historical stock data of 10 companies
• Time period: 2020 – Present (~5+ years)
• Source: Yahoo Finance (yfinance)

⚙️ Methodology

Data Collection & Preprocessing

• Collected stock data using yfinance
• Calculated log returns
• Applied MinMax Scaling
• Split into 80% training / 20% testing

LSTM Model (Time-Series Forecasting)

• Built 2-layer LSTM (100 units each)
• Used 30-day look-back window
• Predicted future stock returns

Portfolio Optimization

• Calculated Expected Return, Volatility, Sharpe Ratio
• Used SciPy SLSQP optimization
• Constraints: weights sum = 1, no short selling

📈 Results

Markowitz Optimization

• Expected Return: 13.45%
• Volatility: 20.44%
• Sharpe Ratio: 0.61

LSTM-Based Optimization

• Expected Return: 53.43%
• Risk: 130.99%

ML-based approach increases return potential but also significantly increases risk.

▶️ How to Use

1. Clone the Repository

git clone https://github.com/Mehtab161/Portfolio-Optimization.git cd portfolio-optimization

2. Install Dependencies

pip install numpy pandas matplotlib seaborn yfinance scikit-learn tensorflow scipy plotly

3. Run the Project

jupyter notebook

Open and run: PortfolioOptimizationFinal.ipynb

⚙️ Customize Inputs

• Modify stock list in code
• Change date range
• Adjust LSTM parameters (LOOK_BACK, FORECAST_RANGE)

📊 Output

• Predicted returns using LSTM
• Optimal portfolio weights
• Expected return, volatility, Sharpe ratio
• Visualizations of trends and allocations

📌 Key Learnings

• Time-series forecasting using LSTM
• Portfolio optimization using MPT
• Understanding risk-return trade-offs
• Combining ML with finance concepts

🔗 Future Improvements

• Add more stocks for diversification
• Use advanced models (GRU, Transformers)
• Add backtesting for validation
• Deploy as a web application

🤝 Connect

• LinkedIn: www.linkedin.com/in/mehtab-khan-mk
• GitHub: https://github.com/Mehtab161

---

⭐ Don’t forget to star the repo if you like it!