[4O5-IS-2c] Machine learning

4:45 PM - 5:00 PM JST(7:45 AM - 8:00 AM UTC)

[4O5-IS-2c-06] Predicting negative news coverage from economic networks using GNNs

〇Sean Brown^1,2, Takayuki Mizuno¹ (1. National Institute of Informatics, 2. Sokendai)

work-in-progress

Keywords:

Graph Neural Networks,Network Analysis,Computational Social Science

Negative reporting that targets companies can have a large impact on the stock market. In this study, we propose a method for predicting negative news from a combination of supply chain, stock ownership, and negative media data. Negative news affects not only the company directly under scrutiny, but can also spread to other firms through shareholdings or other relationships. To investigate this hypothesis, we propose a predictive model using graph neural networks (GNNs) that will utilize the structure of economic networks. The model takes network structure, company attributes, and negative news coverage as inputs and predicts the probability of future negative news coverage. By comparing results to a baseline that does not take network structure into account, we will be able to determine the importance of network structure in making predictions. This research will demonstrate the possibility of predicting risks by considering the structure of interbusiness networks and shows the applicability of computational social science and AI to economic risk analysis.