A DEEP LEARNING–OPTIMIZED HYBRID ARCHITECTURE FOR PREDICTIVE STOCK MARKET MODELING USING SENTIMENT INDICATORS

Abstract

Because of the inherent volatility, nonlinearity, and the impact of both quantitative and
qualitative elements, accurately predicting the
stock market is still a difficult task. In order to
improve forecasting accuracy, this research
suggests a hybrid architecture optimised for
deep learning that combines historical stock data
with market sentiment indices. The model
combines sophisticated deep learning
algorithms, like Long Short-Term Memory
(LSTM) networks and Convolutional Neural
Networks (CNN), for temporal and feature
pattern recognition with Natural Language
Processing (NLP) techniques to extract
sentiment scores from financial news and social
media platforms. The hybrid methodology aligns
sentiment-driven insights with conventional
market signals by using feature fusion and
hyperparameter tweaking techniques to
maximise forecast performance. The suggested
model performs noticeably better than baseline
methods in terms of prediction accuracy and
resilience, according to experimental findings on
benchmark datasets. This study highlights the
possibility of combining artificial intelligence
with human behavioural signals to provide more
accurate and timely stock market predictions.