Scholarlink

<p>Cryptocurrency has played a major role in the growth of digital currency exchange in the current financial environment. The study estimates the values of three cryptocurrencies—Litecoin, Ethereum, and Monero—using deep learning techniques. “Long Short-Term Memory” (LSTM) and “Gated Recurrent Unit” (GRU). Long-term dependencies in sequential data are ideal for recurrent neural networks like LSTM and GRU. The study also used the Direction Algorithm (DA), which takes Bitcoin as a parent coin, to determine its direction, which influences the price of other cryptocurrencies, and Change Point Detection (CPD) approaches, which use Pruned Exact Linear Time (PELT) to detect rapid price variations. In addition, this study uses the Vader algorithm for news sentiment analysis to increase forecast accuracy. Combining Vader, the direction algorithm, and PELT gives the study more depth. The standard LSTM and GRU, as well as the proposed optimized (Vader + direction + PELT) LSTM and GRU, are tested on all three cryptocurrencies for prediction. The proposed LSTM predicted the price of Monero, Ethereum, and Litecoin with an error rate of MAE 1.53, 18.84, and 1.66, respectively. Whereas the proposed GRU model achieved MAE values of 1.11, 16.10, and 1.57 for Monero, Ethereum, and Litecoin price predictions respectively. The obtained results indicate that both the proposed optimized model of LSTM and GRU demonstrated higher accuracy in price prediction compared to standard LSTM and standard GRU models.</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

Optimized Deep Learning Framework for Cryptocurrency Price Prediction

  • 期刊论文

  • 发表日期：

    2025年1月7日

查看全文
  • Rudresh Shirwaikar,
  • Sagar Naik,
  • Abiya Pardeshi,
  • Sailee Manjrekar,
  • Yash Shetye,
  • Siddhesh Dhargalkar,
  • Ritvik Madkaikar

摘要

Cryptocurrency has played a major role in the growth of digital currency exchange in the current financial environment. The study estimates the values of three cryptocurrencies—Litecoin, Ethereum, and Monero—using deep learning techniques. “Long Short-Term Memory” (LSTM) and “Gated Recurrent Unit” (GRU). Long-term dependencies in sequential data are ideal for recurrent neural networks like LSTM and GRU. The study also used the Direction Algorithm (DA), which takes Bitcoin as a parent coin, to determine its direction, which influences the price of other cryptocurrencies, and Change Point Detection (CPD) approaches, which use Pruned Exact Linear Time (PELT) to detect rapid price variations. In addition, this study uses the Vader algorithm for news sentiment analysis to increase forecast accuracy. Combining Vader, the direction algorithm, and PELT gives the study more depth. The standard LSTM and GRU, as well as the proposed optimized (Vader + direction + PELT) LSTM and GRU, are tested on all three cryptocurrencies for prediction. The proposed LSTM predicted the price of Monero, Ethereum, and Litecoin with an error rate of MAE 1.53, 18.84, and 1.66, respectively. Whereas the proposed GRU model achieved MAE values of 1.11, 16.10, and 1.57 for Monero, Ethereum, and Litecoin price predictions respectively. The obtained results indicate that both the proposed optimized model of LSTM and GRU demonstrated higher accuracy in price prediction compared to standard LSTM and standard GRU models.