Pilot Study on the Implementation of Machine Learning Algorithms Using Sentiment Analysis in Financial Markets

Abstract

The fast-growing financial market comprises a vast expanse of quantitative and qualitative insights. The earlier financial prediction models neglect the investor sentiment that can ultimately affect short-term movements in the market. The sentiment-loaded social media, financial news and earnings report predictions data was analysed with the help of Natural Language Processing tools and ML model to predict the stock market trend. This investigation explores whether coupling machine learning with sentiment analysis could meaningfully forecast market fluctuations. Specifically, the study examined whether support vector machines, random forests, and deep learning architectures such as long short-term memory and convolution neural networks - when bolstered by sentiment scoring - might substantially enhance predictive prowess. The study adds to the expanding field of sentiment-aware financial forecasting and offers insights to effective implementation strategies. This initial investigation explores the feasibility of integrating machine learning (ML) algorithms with sentiment analysis to predict fluctuations in financial markets. A lexicon-based method (VADER) and logistic regression is used to analyze sentiment from Twitter messages and financial news titles for a period of six months. Initial findings reveal a moderate predictive accuracy of 65%, indicating the promise of sentiment-driven models despite obstacles like data noise. This research highlights the necessity for more refined models and hybrid strategies in future studies.