Scholarlink

<p>The academic literature lacks a thorough understanding of the intricate architecture of the cryptocurrency asset and derivatives markets, distinguishing it from other asset classes. Its fragmentation, with many self-regulating, centralized and decentralized exchanges functioning in mobile tax havens to evade inspection of their operations and earnings, is a significant distinction from traditional markets. In this work, the historical Bitcoin data is taken from October 2014 to February 2022. We used the multivariate Bidirectional LSTM (Bi-LSTM) model with the <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(TensorFlow\)</EquationSource> </InlineEquation> and <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(Keras\)</EquationSource> </InlineEquation> libraries. The Bi-LSTM model improves the model’s capacity to identify patterns and relationships in complicated, nonlinear time series data, such as cryptocurrency price fluctuations; they are useful for forecasting the volatility of Bitcoin. The model gives a Root Mean Square Error (RMSE) score of 0.0425 and a Root Mean Squared Percentage Error (RMSPE) of 0.1106, which is quite minimal. Lower RMSE and RMSPE values indicate better model performance, as the predicted values are closer to the actual values. Using multivariate bidirectional pattern learning, Bi-LSTM extract attributes from the time-series data. This is very useful when forecasting frequent shifts like the price swings of Bitcoin, which are impacted by both past and future occurrences.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Bitcoin Volatility Prediction: An Analysis of Bi-LSTM Approach

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Nrusingha Tripathy,
  • Sarbeswara Hota,
  • Debabrata Singh,
  • Subrat Kumar Nayak

摘要

The academic literature lacks a thorough understanding of the intricate architecture of the cryptocurrency asset and derivatives markets, distinguishing it from other asset classes. Its fragmentation, with many self-regulating, centralized and decentralized exchanges functioning in mobile tax havens to evade inspection of their operations and earnings, is a significant distinction from traditional markets. In this work, the historical Bitcoin data is taken from October 2014 to February 2022. We used the multivariate Bidirectional LSTM (Bi-LSTM) model with the \(TensorFlow\) and \(Keras\) libraries. The Bi-LSTM model improves the model’s capacity to identify patterns and relationships in complicated, nonlinear time series data, such as cryptocurrency price fluctuations; they are useful for forecasting the volatility of Bitcoin. The model gives a Root Mean Square Error (RMSE) score of 0.0425 and a Root Mean Squared Percentage Error (RMSPE) of 0.1106, which is quite minimal. Lower RMSE and RMSPE values indicate better model performance, as the predicted values are closer to the actual values. Using multivariate bidirectional pattern learning, Bi-LSTM extract attributes from the time-series data. This is very useful when forecasting frequent shifts like the price swings of Bitcoin, which are impacted by both past and future occurrences.