Bitcoin Price Dynamics Analysis Using Modified Multiscale Entropy and Multiscale Poincaré Plots

• Main Authors: Lin, Keng-Hsi, 林耕希
• Other Authors: Yu, Hsin-Chun
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/4m5962

碩士 === 東海大學 === 資訊管理學系 === 107 === Bitcoin, a cryptocurrency based on the application of blockchain technology, has attracted great attention from the global financial market and media for its novel concept different from traditional currencies. Bitcoin has become not only an instrument for payment, investment or risk aversion in the financial market, but also an important case of asset digitization in the context of the boom of digital finance technology. The purpose of the study was mainly to analyze the complexity of Bitcoin trading market, to facilitate understanding whether there is meaningful structure in the dynamic changes of Bitcoin market trading price and studying whether the occurrence of significant events in the world economy and the history of Bitcoin will affect Bitcoin market trading price. In the study, Bitcoin price data during the period from 31 July 2010 to 31 July 2018 were used as samples, and changes in Bitcoin market trading price were regarded as a non-linear time series. Bitcoin trading price data were a short-term time series, therefore, modified multiscale entropy was used in the study to analyze the complexity of Bitcoin trading market and compare with white noise and pink noise. In addition, multiscale poincaré plots were used in the study to implement graphic analysis to the time series of Bitcoin trading price, so as to understand the self-similarity herein. However, the results of data analysis using modified multiscale entropy show that the complexity of Bitcoin market is very low, which indicates that Bitcoin market is similar to white noise and is a relatively random time series. Moreover, the results of analysis using multiscale poincaré plots also show that the self-similarity of Bitcoin market price is very low, and is a random time series similar to white noise. Therefore, it has been proven that the results obtained by us applying such two methods are consistent. It is also found in another study that, the significant events occurred in the history will have an impact on the liquidity of Bitcoin market. Overall, the study shows that changes and movements of Bitcoin trading price is a time series featuring a low complexity, a high randomness, a low self-similarity and also a low future predictability. In addition, both Bitcoin market and trading price are prone to be affected by events.