Title: Information extraction from the GDELT database to analyse the European sovereign bond market
Authors: Luca Tiozzo Pezzoli - JRC European Commission (Italy) [presenting]
Sergio Consoli - National Research Council of Italy (CNR) (Italy)
Elisa Tosetti - University of Venice (Italy)
Abstract: A set of news-based indicators are extracted and used to forecast future behaviour of the sovereign bond yield spread in Italy. We use a big, open-source, news-level database known as Global Database of Events, Language and Tone (GDELT) and extract a large number of variables capturing daily variations in the emotional content of news economic and political events, as well as topics popularity, and use these as proxies for market investor's expectations and behaviour. To this end, we adopt a recurrent network model with feedback connections, known as DeepAR, in combination with a number of approaches for variable reduction, and forecast yield spread at different quantiles. Results show good performance of our methodology for the forecasting of the Italian sovereign bond market using the information extracted from GDELT and a deep Long Short-Term Memory Network opportunely trained and validated with a rolling window approach to best accounting for non-linearities in the data.