NASDAQ 100 Index Prediction Using LSTM And Sentiment Analysis

Stock market prediction constitutes a critical endeavor that significantly influences global economic activities. In the transition from a decade-long bull market characterized by sustained growth to a phase marked by intensified stock price volatility, the establishment of an accurate price forecasting model becomes indispensable. Investors increasingly rely on such models to navigate market uncertainties, mitigate investment risks, and capitalize on emergent opportunities within the financial domain. In this study, the closing price of the NASDAQ 100 index is predicted by a single-layer Long Short-Term Memory (LSTM) network, which takes into account historical pricing data, technical indicators, macroeconomic information, and sentiment scores derived from financial news headlines. The study utilizes three models-BERT, FinBERT-TextCNN, and BBiLSTM-Attention-to extract sentiment scores of varying accuracies from financial news headlines. Additionally, standard metrics such as Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are used to evaluate the predictive performance of LSTM models across varying levels of sentiment score precision. The research explores the question of whether incorporating more precise sentiment scores can enhance the model's predictive performance.