Presently,there are over 500 cryptocurrency exchanges worldwide and more than 19,700 different cryptocurrencies.Despite the fact that cryptocurrency trading is possible via private peer-to-peer transactions,more than ...Presently,there are over 500 cryptocurrency exchanges worldwide and more than 19,700 different cryptocurrencies.Despite the fact that cryptocurrency trading is possible via private peer-to-peer transactions,more than 90%of trading occurs on organised exchanges,which provide convenience and liquidity.However,centralised cryptocurrency exchanges are regarded as high-value targets by criminals and are often victims of cyberattacks.In this paper,we investigate the risk of cryptocurrency exchange closures and develop predictive models to forecast which markets will close down and which ones will remain active using publicly available data.Our models perform well and reach a high level of classification accuracy.Exchange trading volume,the availability of public information on exchange staff,exchange lifetime,and several cybersecurity features are identified as key attributes in predicting exchange closures.Nevertheless,our models do not account for all sources of risk,e.g.,potential fraud and mismanagement of client funds committed by the exchanges themselves,and market participants are encouraged to carefully consider where and how they store their digital assets.展开更多
文摘Presently,there are over 500 cryptocurrency exchanges worldwide and more than 19,700 different cryptocurrencies.Despite the fact that cryptocurrency trading is possible via private peer-to-peer transactions,more than 90%of trading occurs on organised exchanges,which provide convenience and liquidity.However,centralised cryptocurrency exchanges are regarded as high-value targets by criminals and are often victims of cyberattacks.In this paper,we investigate the risk of cryptocurrency exchange closures and develop predictive models to forecast which markets will close down and which ones will remain active using publicly available data.Our models perform well and reach a high level of classification accuracy.Exchange trading volume,the availability of public information on exchange staff,exchange lifetime,and several cybersecurity features are identified as key attributes in predicting exchange closures.Nevertheless,our models do not account for all sources of risk,e.g.,potential fraud and mismanagement of client funds committed by the exchanges themselves,and market participants are encouraged to carefully consider where and how they store their digital assets.
