Stock Portfolio Optimization Using a Combined Approach of Multi Objective Grey Wolf Optimizer and Machine Learning...

Area of Science:

• Quantitative Finance
• Computational Finance
• Investment Management

Background:

• Optimizing stock portfolios is crucial for maximizing returns and minimizing risk in financial markets.
• The Tehran Stock Exchange presents unique challenges and opportunities for portfolio optimization strategies.

Purpose of the Study:

• To optimize stock portfolios of active companies on the Tehran Stock Exchange using forecasted prices.
• To evaluate the effectiveness of combined filtering methods and optimization algorithms for portfolio management.

Main Methods:

• Employed technical analysis indicators (ROC, SMA, EMA, WMA, MACD), Markov Chains, and Machine Learning (Random Forest, Support Vector Machine) for filtering.
• Utilized the mean-variance (M-V) model for portfolio optimization, selecting 85 companies from 480 listed.
• Applied multi-objective optimization algorithms: Multi-Objective Grey Wolf Optimizer (MOGWO) and Non-dominated Sorting Genetic Algorithm II (NSGA II).

Main Results:

• Support Vector Machine (SVM) showed lower correlation error than Random Forest for price prediction.
• Filtering significantly decreased transaction risk and increased stock portfolio returns.
• Simultaneous application of two filtering methods led to a slight increase in portfolio returns and a rise in risk.
• MOGWO algorithm achieved a 133.13% stock return rate with 3.346% risk, outperforming NSGA II's 107.73% return with 1.459% risk.

Conclusions:

• SVM is a preferred method for stock price prediction in this context due to its lower error rate.
• Filtering strategies are effective in enhancing stock portfolio performance by reducing risk and increasing returns.
• The MOGWO algorithm demonstrates superior efficiency and effectiveness in optimizing stock portfolios compared to NSGA II.