Using ML and Explainable AI to understand the interdependency networks between classical economic indicators and crypto-markets

Todorovska, Ana; Peshov, Hristijan; Rusevski, Ivan; Vodenska, Irena; Chitkushev, Lubomir T.; Trajanov, Dimitar

doi:10.1016/j.physa.2023.128900

Cited by 3 publications

(1 citation statement)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition to employing econometric models, researchers leveraged Machine Learning (ML) and explainable Artificial Intelligence (AI) to explore the intricate relationship between macroeconomic indicators, financial markets, and cryptocurrencies [53]. Their methodology involved the analysis of both structured and unstructured data, encompassing numerical and text data from sources like news and social networks.…”

Section: Literature Reviewmentioning

confidence: 99%

Exploring the Dynamics of Brent Crude Oil, S&P500 and Bitcoin Prices Amid Economic Instability

Bâra,

Georgescu,

Oprea

et al. 2024

IEEE Access

View full text Add to dashboard Cite

In this paper, we mainly investigate three variables from the price volatility point of view: Brent crude oil, S&P500 and Bitcoin (BTCUSD), aiming to underline the impact of price volatility. Brent crude oil accounts for two-thirds of the oil market. Its price volatility has a significant impact on environmental, transportation, mobility, economic and social aspects that affect sustainability. This paper conducts an extensive examination of the forecasting capabilities of various GARCH (Generalized Autoregressive Conditional Heteroskedasticity) models, identifying the most suitable GARCH model for estimating Value at Risk (VaR) for Brent crude oil price. The assessment of VaR for different GARCH models is carried out using Kupiec's Probability of Failure (POF) test and Christoffersen's test. This study leverages Brent crude oil data spanning from 2019 to 2023. Additionally, to prove the robustness of the GARCH models, we further consider the West Texas Intermediate (WTI) and Dubai oil prices that are the dominant in the U.S and Asian market. The investigation identifies the TGARCH(1,1) Skewed Student model as the optimal choice among 9 models considered for VaR estimation. The results show that TGARCH Skewed Student model surpasses the other models in the study, proving its superiority in forecasting Brent crude oil price volatility and facilitating VaR estimation. A VaR of 0.044 with a 95% confidence level means that there is a 95% chance that the portfolio will not lose more than 4.4% of its value. By incorporating skewness in addition to volatility asymmetry, the Skewed GARCH-type models provide a more realistic representation of the underlying return distribution. Furthermore, the most appropriate GARCH-type model for WTI crude oil is EGARCH(1,1) Skewed Student, with a VaR coverage of 0.39. The most appropriate GARCH-type model for Dubai crude oil is TGARCH(1,1) Skewed Student, with a VaR coverage of 0.17. Both WTI oil and Dubai crude oil have a coverage that exceeds 5%, implying a more conservative approach to estimating potential losses. Furthermore, the unidirectional causalities BTCUSD→BRENT and BTCUSD→S&P500 are identified. The results of the current research have practical implications for both importing and exporting countries, policy makers and investors. For companies in the oil sector, VaR informs operational decisions, such as production levels, capital expenditure and inventory management, by providing insights into market risk. Moreover, understanding the risks associated with oil aids in long-term strategic planning.INDEX TERMS backtesting, Bitcoin, Brent crude oil, GARCH model, price volatility, S&P500

show abstract

Section: Literature Reviewmentioning

confidence: 99%