EXPLORING CRYPTOCURRENCY TIME SERIES DATA USING TOPOLOGICAL DATA ANALYSIS
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Malaysia Sarawak
Abstract
Description
Cryptocurrencies, especially Bitcoin, are widely known for their significant volatility and unpredictable price movements, complicating efforts to analyze and predict their behaviour. Traditional time-series models often fail to capture the complex nonlinear dynamics and irregularities in Bitcoin's pricing data, leading to an incomplete understanding of its market trends. This study introduces topological data analysis as a new approach to address these issues. The implementation involved transforming Bitcoin’s daily price data into point clouds using Takens embedding and analyzing them with persistent homology. Persistence diagrams were generated for four market phases, which are flat, stable trend, bullish and volatile. The results showed that stable periods produced simpler topological patterns, while volatile periods produced more persistent loops, reflecting complex and repeating behaviours. These findings highlight how topological data analysis reveals structural differences across market conditions. In conclusion, topological data analysis offers a powerful alternative to traditional analysis by uncovering hidden patterns in time series data, providing deeper insights into cryptocurrency price dynamics.