Сравнителен анализ на статистически и AI базирани модели при оценка на влиянието на новини върху пазарната цена на Bitcoin
Comparative Analysis of Statistical and AI-Based Models in Assessing the Impact of News on the Market Price of Bitcoin

Author(s): Nikita Shvetsov
Subject(s): Social Sciences, Economy, Sociology, Policy, planning, forecast and speculation, Social Informatics, Financial Markets, ICT Information and Communications Technologies, Socio-Economic Research
Published by: Университет за национално и световно стопанство (УНСС)
Keywords: news impact; sentiment analysis; machine learning; price movement forecasting; ai-based models
Summary/Abstract: This study examines the impact of news sentiment and Google Trends indices on the price of Bitcoin by applying both classical and modern statistical models. Various methods are employed for the analysis, including linear regression, XGBoost, and LSTM. The model is based on temporal dependencies between the sentiment index, trend interest, and market movement. The results indicate that the regression model with the variables NSI and Trend partially explains price movements, with NSI showing a positive effect. XGBoost demonstrates high predictive accuracy (R² ≈ 0.78), while LSTM captures nonlinear and deep dependencies in the data. The analysis reveals that the effect of news is highly short-term, making classical ARIMA models less applicable. The study contributes to a better understanding of behavioral factors and highlights the role of AI models in economic forecasting.

• Details
• Contents