TradeWise: Towards Context-Aware Stock Market Predictions with Sentiment and Political Insights

Abstract

Our study includes the preprocessing of a year’s worth of historical data, which combines traditional financial metrics with sentiment scores derived from financial news and indices reflecting the prevailing political climate. This enriched dataset is employed to train four different machine learning algorithms: a Hybrid, a Random Forest Model (RFM), a Support Vector Machine (SVM), and a K-Nearest Neighbors (KNN) model. Our results indicate that the inclusion of sentiment and political data contributes positively to the performance of all test models, with significant enhancements noted particularly in precision and F1-scores. Our novel approach suggests that sentiment and political insights, when processed and integrated effectively, offer substantial predictive value that could refine the accuracy of financial prediction models. This enhanced performance underscores the potential of combining qualitative analyses with quantitative data to create more robust predictive models in finance.