Citation

Jasbi, Ali and Subbarao, Anusuyah and Kumar, Kiran Raj Raj (2026) Predicting Bitcoin Price Movements Using Machine Learning: An Application of XGBoost on Second-level Market Data and Financial Indicators. In: 2026 International Conference on Cognitive Systems and Computer Interaction, ICoSCI 2026, 15 - 16 January 2026, Kuala Terengganu.

Text IEEE Xplore Full-Text PDF_2.pdf - Published Version Restricted to Repository staff only Download (598kB)

Abstract

The volatile and decentralized nature of cryptocurrency markets presents significant challenges in accurate price prediction, particularly for assets like Bitcoin. This study explores the application of the XGBoost machine learning algorithm to predict Bitcoin price movements across multiple time intervals using second-level market data collected from Binance, along with curated financial indicators. Key features such as MACD, RSI, EMA, Bollinger Bands, and order book imbalances are incorporated to capture market dynamics. The model demonstrates an average Root Mean Square Error (RMSE) of 0.03 for shorter intervals, emphasizing its robustness in high-frequency trading environments. The multi-interval feature importance analysis reveals the differential relevance of liquidity metrics and momentum indicators across short- and long-term predictions, offering actionable insights as part of a cognitive decision-support framework for traders. While this study focuses on XGBoost due to its interpretability, future research will extend this framework through comparative analysis with deep learning models, including LSTM and GRU architectures, and application to additional cryptocurrencies.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Cryptocurrency prediction, XGBoost, financial indicators
Subjects:	H Social Sciences > HG Finance
Divisions:	Faculty of Management (FOM)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	05 Jun 2026 07:55
Last Modified:	08 Jun 2026 09:25
URII:	http://shdl.mmu.edu.my/id/eprint/16063

Downloads

Edit (login required)