Predict stock price fluctuations using Realized Volatility, CEEMDAN, LSTM models

Abstract

In today’s rapidly evolving financial markets, the fluctuation of stock prices has a significant impact on the decision-making of investors. To better understand and predict these price movements, this paper proposes an integrated approach aimed at enhancing the accuracy of predictions. Initially, this paper analyzes the characteristics of price fluctuations in the U.S. stock market and discusses their influence on investor decision-making. Building on this foundation, a new forecasting model is introduced in this paper, which combines various advanced time series analysis techniques, including Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), Long Short-Term Memory networks (LSTM), Autoregressive (AR), and High-Order Autoregressive (HAR) models. By comparing the performance of these models under different market conditions, this paper aims to assess their effectiveness and reliability in predicting stock prices. The research results indicate that the combination of these models can significantly improve the accuracy of predictions, providing investors with a more reliable decision-making tool. Furthermore, this paper also explores the applicability and limitations of these models under various market conditions, offering valuable insights for future research and practice.