Stock price prediction is a challenging problem due to high volatility, non-linearity, and influence of external factors such as public sentiment and global events. Traditional time-series models rely mainly on historical price data and often fail to capture market psychology. In recent years, deep learning models such as Generative Adversarial Networks (GANs) have shown promising results in complex data generation tasks, but their application to financial time-series forecasting remains limited.
This project proposes a novel stock price forecasting model using a Generative Adversarial Network (GAN) that integrates historical stock prices, technical indicators, and Twitter sentiment analysis. Amazon (AMZN) stock data is used as a case study. Twitter sentiment scores are computed using the VADER sentiment analyzer, and multiple technical indicators such as Moving Averages, Bollinger Bands, and MACD are incorporated. Experimental results show that combining sentiment and technical indicators improves prediction accuracy compared to using historical prices alone.
