DEEP LEARNING-BASED STOCK FORECASTING SYSTEM ENHANCED WITH SENTIMENT ANALYSIS

Abstract

Stock markets fluctuate continuously under the influence of numerous factors, including past trends, eco- nomic indicators, public mood, and news reports. Due to these irregular and rapidly changing patterns, achieving precise stock price predictions is extremely complex. This study proposes an integrated fore- casting framework that combines Long Short-Term Memory (LSTM) neural networks with sentiment analysis to enhance prediction performance. Historical stock data and financial news are combined to create enriched datasets featuring technical indicators and sentiment scores. The model leverages deep learning for sequential pattern recognition and NLP for sentiment extraction. The findings reveal that merging textual sentiment with traditional technical metrics significantly refines predictive accuracy, offering dependable support for financial analysts and investors.