Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis

Bitcoin has recently grown to prominence as a decentralized digital currency, attracting significant interest for its potential transformation of the financial market. Forecasting Bitcoin's price is crucial for investors, traders, and academics, given the currency's inherent volatility, wh...

Full description

Saved in:
Bibliographic Details
Main Authors: Rachid Bourday, Issam Aatouchi, Mounir Ait Kerroum, Ali Zaaouat
Format: Article
Language:English
Published: Ital Publication 2024-12-01
Series:HighTech and Innovation Journal
Subjects:
Online Access:https://hightechjournal.org/index.php/HIJ/article/view/641
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846100191658115072
author Rachid Bourday
Issam Aatouchi
Mounir Ait Kerroum
Ali Zaaouat
author_facet Rachid Bourday
Issam Aatouchi
Mounir Ait Kerroum
Ali Zaaouat
author_sort Rachid Bourday
collection DOAJ
description Bitcoin has recently grown to prominence as a decentralized digital currency, attracting significant interest for its potential transformation of the financial market. Forecasting Bitcoin's price is crucial for investors, traders, and academics, given the currency's inherent volatility, which makes accurately predicting future prices challenging. This article aims to provide a comprehensive and comparative analysis of Deep Learning Forecasting Models in order to predict Bitcoin prices in the short and medium terms: Transformer with XGBoost, Transformer with ANN, Transformer with LSTM, and Transformer with SVR. This study is the first to explore the effectiveness of transformer-based architectures, particularly focusing on feature extraction, in complex financial market predictions. Therefore, we trained these models using historical Bitcoin data from 2016 to 2023 and evaluated their performance on a test dataset. Our experiments demonstrate that the Transformer with the XGBoost model outperforms the baseline models, achieving a Mean Absolute Error (MAE) of 0.011 and a Root Mean Squared Error (RMSE) of 0.018. Our findings suggest that the use of advanced deep learning techniques effectively manages the complexities of the cryptocurrency market, offering significant improvements over traditional methods and guiding investors in the cryptocurrency markets.   Doi: 10.28991/HIJ-2024-05-04-013 Full Text: PDF
format Article
id doaj-art-ace82e0fff7a4c2f99c1b45e2e1d1c53
institution Kabale University
issn 2723-9535
language English
publishDate 2024-12-01
publisher Ital Publication
record_format Article
series HighTech and Innovation Journal
spelling doaj-art-ace82e0fff7a4c2f99c1b45e2e1d1c532024-12-30T12:24:57ZengItal PublicationHighTech and Innovation Journal2723-95352024-12-01541055106710.28991/HIJ-2024-05-04-013229Cryptocurrency Forecasting Using Deep Learning Models: A Comparative AnalysisRachid Bourday0Issam Aatouchi1Mounir Ait Kerroum2Ali Zaaouat3Computer Science Research Laboratory, Faculty of Sciences, IBN Tofail University, Kenitra,Computer Science Research Laboratory, Faculty of Sciences, IBN Tofail University, Kenitra,Computer Science Research Laboratory, Faculty of Sciences, IBN Tofail University, Kenitra,Computer Science Research Laboratory, Faculty of Sciences, IBN Tofail University, Kenitra,Bitcoin has recently grown to prominence as a decentralized digital currency, attracting significant interest for its potential transformation of the financial market. Forecasting Bitcoin's price is crucial for investors, traders, and academics, given the currency's inherent volatility, which makes accurately predicting future prices challenging. This article aims to provide a comprehensive and comparative analysis of Deep Learning Forecasting Models in order to predict Bitcoin prices in the short and medium terms: Transformer with XGBoost, Transformer with ANN, Transformer with LSTM, and Transformer with SVR. This study is the first to explore the effectiveness of transformer-based architectures, particularly focusing on feature extraction, in complex financial market predictions. Therefore, we trained these models using historical Bitcoin data from 2016 to 2023 and evaluated their performance on a test dataset. Our experiments demonstrate that the Transformer with the XGBoost model outperforms the baseline models, achieving a Mean Absolute Error (MAE) of 0.011 and a Root Mean Squared Error (RMSE) of 0.018. Our findings suggest that the use of advanced deep learning techniques effectively manages the complexities of the cryptocurrency market, offering significant improvements over traditional methods and guiding investors in the cryptocurrency markets.   Doi: 10.28991/HIJ-2024-05-04-013 Full Text: PDFhttps://hightechjournal.org/index.php/HIJ/article/view/641forecastingdeep learningmachine learningtransformerlstmxgboost.
spellingShingle Rachid Bourday
Issam Aatouchi
Mounir Ait Kerroum
Ali Zaaouat
Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
HighTech and Innovation Journal
forecasting
deep learning
machine learning
transformer
lstm
xgboost.
title Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
title_full Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
title_fullStr Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
title_full_unstemmed Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
title_short Cryptocurrency Forecasting Using Deep Learning Models: A Comparative Analysis
title_sort cryptocurrency forecasting using deep learning models a comparative analysis
topic forecasting
deep learning
machine learning
transformer
lstm
xgboost.
url https://hightechjournal.org/index.php/HIJ/article/view/641
work_keys_str_mv AT rachidbourday cryptocurrencyforecastingusingdeeplearningmodelsacomparativeanalysis
AT issamaatouchi cryptocurrencyforecastingusingdeeplearningmodelsacomparativeanalysis
AT mouniraitkerroum cryptocurrencyforecastingusingdeeplearningmodelsacomparativeanalysis
AT alizaaouat cryptocurrencyforecastingusingdeeplearningmodelsacomparativeanalysis