Scholarlink

A virtual or digital currency, cryptocurrency, is based on blockchain technology and uses encryption to ensure security. Currently, over 2000 different coin types are available on the cryptocurrency market, with a massively unequal number of transactions and circulation. Anticipating the inclination of cryptocurrency prices to fluctuate is essential since the investment risk attached to them is greater than that of conventional goods. Four different models were utilized in this work to address the limitations of traditional production forecasting: Long Short-Term Memory (LSTM), Facebook Prophet (FB-Prophet), Silverkite, and Bidirectional LSTM. The FB Prophet and Silverkite both support the exponential smoothing method, where the LSTM and Bi-LSTM are the deep learning models. Silverkite is the main algorithm used in the Python library Graykite by LinkedIn. We examined the models using historical Bitcoin data over the previous nine years, from January 2012 to March 2021. The Bi-LSTM model provides a root mean squared error (RMSE) score of 3.415 and a mean absolute error (MAE) score of 7.539. The Bi-LSTM model detects the variations that might attract attention and prevent potential issues.

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

A Comparison of Interdependent Deep Learning Models and Exponential Smoothing Method for Predicting Bitcoin Price

  • 章节

  • 发表日期：

    2025年

查看全文
  • Nrusingha Tripathy,
  • Sarbeswara Hota,
  • Debahuti Mishra,
  • Meera Dash,
  • Soumyarashmi Panigrahi,
  • Subrat Kumar Nayak

摘要

A virtual or digital currency, cryptocurrency, is based on blockchain technology and uses encryption to ensure security. Currently, over 2000 different coin types are available on the cryptocurrency market, with a massively unequal number of transactions and circulation. Anticipating the inclination of cryptocurrency prices to fluctuate is essential since the investment risk attached to them is greater than that of conventional goods. Four different models were utilized in this work to address the limitations of traditional production forecasting: Long Short-Term Memory (LSTM), Facebook Prophet (FB-Prophet), Silverkite, and Bidirectional LSTM. The FB Prophet and Silverkite both support the exponential smoothing method, where the LSTM and Bi-LSTM are the deep learning models. Silverkite is the main algorithm used in the Python library Graykite by LinkedIn. We examined the models using historical Bitcoin data over the previous nine years, from January 2012 to March 2021. The Bi-LSTM model provides a root mean squared error (RMSE) score of 3.415 and a mean absolute error (MAE) score of 7.539. The Bi-LSTM model detects the variations that might attract attention and prevent potential issues.