PREDICTION OF NASDAQ FUTURES INDEX BASED ON SENTIMENT ANALYSIS AND MACRO DATA

초록

Stock price prediction remains a challenging task in the financial field, with various models being developed to enhance prediction accuracy. In this study, we investigate the efficacy of three predictive models – Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Extreme Gradient Boosting (XGBoost) – to forecast the Nasdaq 100 futures index. We build these models using financial data up to day D to predict the stock price for the following day (D+1). Additionally, the study incorporates macroeconomic data such as the Consumer Price Index (CPI), interest rates, and sentiment analysis from news articles to improve the model’s performance. We evaluate the models using performance metrics including Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Root Mean Squared Logarithmic Error (RMSLE) and R2 score. Our findings show that integrating macroeconomic and sentiment data improves prediction accuracy significantly. Among the three models, XGBoost (Extreme Gradient Boosting) outperforms both LSTM and GRU, demonstrating superior performance with the lowest error metrics, including a remarkable R2 score of 0.9892 when incorporating macroeconomic and sentiment data. This highlights the importance of incorporating broader market and sentiment data in stock price prediction. © 2025, ICIC International. All rights reserved.

키워드