Enhancing Portfolio Allocation by LSTM Model

Abstract

Accurate forecasting of stock prices and the construction of optimized portfolios are essential for investors in the dynamic technology sector. This paper proposes a comprehensive approach that combines Long Short-Term Memory (LSTM) models with portfolio construction techniques specifically tailored to the stocks of Apple Inc. (AAPL), Meta Platforms Inc. (META), Amazon.com Inc. (AMZN), Microsoft Corporation (MSFT), and NVIDIA Corporation (NVDA) from January 1st ,2018 to May 31st, 2023. By leveraging the sequential nature of historical stock price data(80% of data), LSTM models capture complex patterns and dependencies, enabling more precise predictions of Adjusted Close (Adj Close) prices. Subsequently, the forecasted prices (20% of data)are utilized to construct optimized portfolios that maximize returns and minimize risks within the technology sector using Monte Carlo simulations, efficient frontier analysis, and key risk-return metrics. The overall result of the prediction data is similar to the actual data which implies that the integration of LSTM-based forecasting and portfolio construction provides a robust framework for informed investment decision-making and risk management. And the application of Monte Carlo simulations, efficient frontier analysis, and key risk-return metrics gave us two portfolio allocation options : Minimum Variance model (40% of AAPL,60% of MSFT) and Maximum Sharpe Ratio model ( 47% of META, 53% of NVDA). The evaluation of the two portfolios show that the strategy can significantly beat the SP500 index.