Hierarchical Structure Estimation of Financial Networks and Sentiment Analysis of Financial News Headlines

In financial markets, there have been several crises, and it is often suggested that early signs of these crises can be observed in market structures before they fully emerge. In this study, we apply the one-layer Hierarchical Stochastic Block Model (HSBM) to detect anomalies by analyzing market changes and the underlying spatial structure, providing crucial insights to support investors’ investment decisions. Moreover, understanding and explaining these market shifts is essential, as it enables investors to respond effectively to fluctuations and proactively prepare for similar major changes in the future. To deepen this analysis, we extended to two-layer HSBM, treating the blocks in the lower layer as nodes in the upper layer, using a two-layer approach for cause analysis. Additionally, the sentiment score of financial news headlines is widely regarded as a strong indicator of market conditions. During periods of market prosperity, sentiment scores tend to be more positive, while in downturns, they become more negative. In this study, we analyzed the structural changes in the S&P 100 indices and compared them with the sentiment scores of financial news headlines from the same period. Our findings reveal a correlation between sentiment scores and market structure changes under normal market conditions. However, during the pandemic, discrepancies between the two emerged, suggesting that crisis periods may disrupt this relationship. These insights provide a new perspective for understanding financial market dynamics.