Summarization-enhanced Sentiment Analysis for Stock Market Forecasting

Abstract

The ever changing nature of financial markets pre¬sents significant hurdles for individuals looking to make well informed choices. In recent times delving into the sentiments conveyed within news articles has emerged as a potent tool for predicting market trends. However, the overwhelming volume of financial news data renders manual analysis virtually impractical leading to the adoption of automated summarization techniques. This research introduces an innovative method for pre¬dicting stock market fluctuations by intertwining sentiment analysis with cutting-edge text summarization strate¬gies. Our approach revolves around the utilization of Generative Adve¬rsarial Networks (GANs) for text summarization. By leveraging the Pointer Generator and Coverage Mechanism our model significantly enhances the quality and coherence of the generated summaries resulting in notable achievements with a ROUGE-1 score of 36.3 and a ROUGE-L score of 31.7. Moreover our sentiment model acts as a compass guiding us through the seas of sentiments present in both the initial news articles and their abridged counterparts. Our evaluation of the approach involved immersing ourselves in real-world financial datasets navigating through over 5k news articles and corresponding stock trends. The predictive model, fueled by sentimental scores from these condensed news pieces, notched up an accuracy of 70% and 71% with the Random Forest and Naive¬ Bayes classifiers and an astounding 99.23% using a pretrained transformer. Our approach streamlines the analysis of financial news and improves the accuracy and reliability of stock market predictions. By effectively summarizing and extracting sentiment information from large volumes of textual data, our method provides a valuable tool for navigating the complexities of financial markets. One can benefit from concise, sentiment-infused summaries that capture the essence of market-relevant information, enabling more informed decision-making and potentially improving investment strategies.