The Study on the Application of Machine Learning Algorithms for Stock Prices Prediction During Special Periods

Stocks are a favored investment channel, with their fluctuations closely linked to macroeconomic conditions. In recent years, global events such as COVID-19, the Russo-Ukrainian war, and the Israeli-Palestinian conflict have caused significant volatility in international stock markets. This paper investigates the application of machine learning algorithms for stock price prediction during these turbulent times. Models such as Random Forest, Support Vector Machine, Linear Regression, Convolutional Neural Networks, Artificial Neural Networks, and Long Short-Term Memory Networks are frequently employed during such periods. Many of these models integrate policy and news indicators that reflect social changes and investor sentiment, helping improve prediction accuracy in times of high uncertainty. Additionally, some models introduce methods for optimizing hyperparameters to enhance forecasting performance further. However, existing machine learning models face challenges such as low interpretability, limited applicability, and high sensitivity to external factors. These issues often lead to reduced investor confidence, increased training costs, and inconsistent results across different stocks and special periods. Looking forward, the integration of new approaches, such as expert systems, alongside traditional machine learning methods, could help mitigate these challenges and improve prediction outcomes in future periods of economic uncertainty.