A Comprehensive Review on AI-Powered Stock Market Prediction Using Sentiment Analysis and Deep Learning Models

Abstract

With the advent of the artificial intelligence (AI), natural language processing (NLP) and deep learning models, the financial stock market prediction has changed radically. The classical techniques of forecasting were based upon numbers of indicators such as price history, candlestick patterns and financial measurements but according to some of the research studies, market sentiments that are derived through financial news, tweets, reports and social behaviours are gigantic contributors to the fluctuation of the prices of the stocks. The more promising results with respect to both the trend forecasting and the risk estimation came from the models such as LSTM, GRU, BERT, FinBert, transformers, multimodal fusion network, and sentiment-aware hybrid architecture, etc. The main technological breakthroughs during 2019-2025, its relative strengths, benchmark dataset(s), limitations (e.g. explainable AI, adaptive transformer pipelines, multimodal trading agent, emotion-driven financial analytics), future direction of research etc. are collected together in this review paper. Utilizing sentiment signals along with machine learning models to improve their ability to predict in the future is an area of interest, and this paper aims to provide a detailed perspective to elicit the process of making better informed and accurate stock predictions and relieve stock forecasting systems of much of the burden.