Scholarlink

<p>Portfolio optimization remains a central challenge in finance, demanding accurate stock forecasting and effective risk-return trade-offs. This study introduces a novel hybrid model, the Improved Arithmetic Optimization Algorithm Long Short-Term Memory (IAOA-LSTM), tailored for portfolio selection in the biotechnology and oil &amp; gas sectors. Leveraging a decade-long dataset comprising daily prices of 50 representative stocks, the proposed model integrates the temporal modeling strength of LSTM with the global search capabilities of an enhanced Arithmetic Optimization Algorithm. Comparative analyses against standard Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Grid Search LSTM(GRID-LSTM), Random Search LSTM (RANDOM-LSTM), Genetic Algorithm LSTM (GA-LSTM) models demonstrate IAOA-LSTM’s superior predictive performance, achieving the lowest errors. Stocks were ranked using a reconstruction error-based metric, and the most predictable were selected to construct optimized portfolios. Furthermore, the efficient frontier, derived via Monte Carlo simulations, identified the portfolio with the highest Sharpe ratio at volatility-return coordinates, offering the most favorable risk-adjusted return. Rigorous statistical testing confirms the model’s significant improvement over benchmarks. These findings underscore the potential of IAOA-LSTM in enhancing investment strategies through deep learning (DL) and bio-inspired optimization. By aligning sector-specific dynamics with advanced modeling, this research offers a robust decision-support tool for investors aiming to maximize returns under uncertainty.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

IAOA-LSTM: a hybrid model for stock portfolio optimization

  • 期刊论文

  • 发表日期：

    2026年1月19日

查看全文
  • David Opeoluwa Oyewola,
  • Temidayo Oluwatosin Omotehinwa,
  • Temitope Olubanjo Kehinde,
  • Yahya Hammawa

摘要

Portfolio optimization remains a central challenge in finance, demanding accurate stock forecasting and effective risk-return trade-offs. This study introduces a novel hybrid model, the Improved Arithmetic Optimization Algorithm Long Short-Term Memory (IAOA-LSTM), tailored for portfolio selection in the biotechnology and oil & gas sectors. Leveraging a decade-long dataset comprising daily prices of 50 representative stocks, the proposed model integrates the temporal modeling strength of LSTM with the global search capabilities of an enhanced Arithmetic Optimization Algorithm. Comparative analyses against standard Long Short Term Memory (LSTM), Gated Recurrent Unit (GRU), Grid Search LSTM(GRID-LSTM), Random Search LSTM (RANDOM-LSTM), Genetic Algorithm LSTM (GA-LSTM) models demonstrate IAOA-LSTM’s superior predictive performance, achieving the lowest errors. Stocks were ranked using a reconstruction error-based metric, and the most predictable were selected to construct optimized portfolios. Furthermore, the efficient frontier, derived via Monte Carlo simulations, identified the portfolio with the highest Sharpe ratio at volatility-return coordinates, offering the most favorable risk-adjusted return. Rigorous statistical testing confirms the model’s significant improvement over benchmarks. These findings underscore the potential of IAOA-LSTM in enhancing investment strategies through deep learning (DL) and bio-inspired optimization. By aligning sector-specific dynamics with advanced modeling, this research offers a robust decision-support tool for investors aiming to maximize returns under uncertainty.