Stock Price Prediction Leveraging News Headlines

Predicting stock prices remains a challenging task in the financial and trading industries due to the high volatility and complex dependencies involved. Traditional methods such as ARIMA models have been supplemented by deep learning techniques, including hybrid neural network models, to improve forecasting accuracy. Recent research has explored incorporating diverse data sources, such as sentiment analysis from social media, Google Trends data, and news headlines, to enhance predictive models. This paper introduces a multivariate approach that integrates news headlines with stock price data to predict future stock prices. The challenge of data collection is addressed by scraping news headlines and cleaning the data using trans-former models. The processed news data is then combined with stock price data obtained from Yahoo Finance's API. A stacked multivariate LSTM model is trained on this combined dataset for stock price prediction. While existing literature offers numerous models for time series analysis and some for predicting stock movement, research on incorporating news headlines for stock price prediction remains limited. This paper attempts to fill this gap by leveraging news information to enrich the deep learning model. The paper reviews existing works in time series analysis, describes the data acquisition and processing techniques, explains the LSTM model architecture for stock price prediction with news headlines, presents experimental results, and concludes with insights and future directions.