A0478
Title: Social media information to forecast Bitcoin value: A comparison of vines and graphical models
Authors: Lorenzo Merli - University of Pavia (Italy) [presenting]
Claudia Tarantola - University of Pavia (Italy)
Luciana Dalla Valle - University of Plymouth (United Kingdom)
Silvia Angela Osmetti - Università Cattolica di Milano (Italy)
Abstract: The aim is to enhance Bitcoin price forecasts by leveraging graphical models and vine copulas. By integrating daily Bitcoin prices with Google Trends data, Twitter activity, and sentiment analysis using Bing and Afinn lexicons, the complex relationships within Bitcoin trends are captured. One hundred fourteen (114) daily observations from February to May 2021 are utilized. Mixed graphical models (MGM) and vector autoregressive (VAR) models forecast Bitcoin prices, while ARIMA-GARCH and gamlss models extract residuals for vine copula implementation. Vine models predict Bitcoin prices using a rolling window method. Comparing forecasts with observed data highlights model accuracy, providing a comprehensive view of Bitcoin market dynamics and public sentiment.
