Given the substantial volatility and non-stationarity of cryptocurrency prices, forecasting them has become a complex task within the realm of financial time series analysis. This study introduces an innovative hybrid prediction model, VMD-AGRU-RESVMD-LSTM, which amalgamates the disintegration–integration framework with deep learning techniques for accurate cryptocurrency price prediction. The process begins by decomposing the cryptocurrency price series into a finite number of subseries, each characterized by relatively simple volatility patterns, using the variational mode decomposition (VMD) method. Next, the gated recurrent unit (GRU) neural network, in combination with an attention mechanism, predicts each modal component’s sequence separately. Additionally, the residual sequence, obtained after decomposition, undergoes further decomposition. The resultant residual sequence components serve as input to an attentive GRU (AGRU) network, which predicts the residual sequence’s future values. Ultimately, the long short-term memory (LSTM) neural network integrates the predictions of modal components and residuals to yield the final forecasted price. Empirical results obtained for daily Bitcoin and Ethereum data exhibit promising performance metrics. The root mean square error (RMSE) is reported as 50.651 and 2.873, the mean absolute error (MAE) stands at 42.298 and 2.410, and the mean absolute percentage error (MAPE) is recorded at 0.394% and 0.757%, respectively. Notably, the predictive outcomes of the VMD-AGRU-RESVMD-LSTM model surpass those of standalone LSTM and GRU models, as well as other hybrid models, confirming its superior performance in cryptocurrency price forecasting.
Cryptocurrency prediction is important for a variety of stakeholders, from investors to businesses, as it enables them to make more informed decisions about the future of the digital asset market. This paper delves into the application of deep learning models for two of the most popular cryptocurrencies, Bitcoin and Ethereum, outlining how to effectively implement these methods. Our goal is to perform efficient deep learning structure based on the forecasting models specifically recurrent neural networks, convolutional neural network and long short-term memory to predict the Bitcoin and Ethereum prices. Our results include a comparison of these two cryptocurrencies according to the deep learning methods and their effectiveness in predicting the Bitcoin and Ethereum prices.
In the long run, Bitcoin transaction fees are the only source of revenue for miners. They compete broadly in two main ways: proof of work effort to win blocks; and transaction processing to gather fee rewards into the blocks they win. This paper contributes to existing literature by developing a dynamic model that separates these two functions, and explores implications for aggregate efficiency outcomes. Specifically, when set by free market forces (unrestricted by artificially imposed block size caps), what happens to overall transaction prices and quantities relative to total energy use? When is it worth Stackelberg-leading miners investing in efficiency-improving R&D? What effect does this have on overall efficiency over time? By explicitly separating specialised capital dedicated to SHA256 hashing (for proof of work) from transaction processing capital (for transaction collection and verification), this paper sheds light on these questions. One key conclusion is that miner innovation lowers energy use per transaction over time for elastic enough transaction demand schedules. The more competitors Bitcoin has (existing fiat and data services, and other new Blockchain-based systems), the stronger is this conclusion.
The complicated combination of long-term patterns, short-term seasonality, and uncertainty has made selecting an appropriate forecasting model for cryptocurrency prices challenging. To forecast the future prices of the cryptocurrencies, this study compared three machine learning models: the Seasonal Auto-Regressive Integrated Moving Average (SARIMA), the Long Short-Term Memory (LSTM), and the Facebook (Fb) Prophet for the period of 2017 till 2023. It was discovered that the FbProphet model works well in predicting the daily price forecasts of Bitcoin and Dogecoin with low mean squared error (MSE), using time series datasets from 2017 through 2023 as the training data to anticipate 12 months of unseen data. While predicting the future from 2023 to 2024, Fb Prophet stated that Dogecoin is most likely to remain stable while Bitcoin price is most likely to continue falling. The LSTM model performs better than the other models based on forecasting results.
<p>The objective of the study is to use daily Thai data analysis to strengthen correlations between Bitcoin and conventional asset measurements. The most popular asset prices and indices include gold, oil, the SET50 index, Bitcoin (BTC), Ethereum (ETH), Litecoin (LTC), Ripple (XRP), Dashcoin (DASH), Stellar Lumens (XLM), Binance coin (BNB), and Dogecoin (DOGE). We find a significant correlation between cryptocurrencies and the digital economy using a matrix approach to the Pearson correlation coefficient. With the help of a minimal spanning tree model and random matrix theory, we can determine the shortest route between assets. Yet, as predicted, only a small percentage of the greatest eigenvalues diverge. We are also developing a novel technique to find the SET-50 index. In an investment portfolio during the coronavirus period, alternatives to the gold price and the DOGE may offer possibilities for risk diversification.</p>
Purpose The study was done to review the existing literature available on the theme using a popular technique known as a bibliometric review. The purpose was to explore important bibliometric trends such as geographical distribution of research; the most relevant countries and institutions and important collaboration networks, frequently published authors, the most relevant topics/research domains and relationships among these, average citations or per year, the most relevant sources, top authors’ production, authors’ impact by H index and the progression of important keywords over a period of time. Design/methodology/approach The study analyzed literature published in the English language from 2012 onwards that used the words “cryptocurrency”, “Ethereum” “Bitcoin” along with “investment/s” or “speculation/s” in the Title/ABS/KEY. A specialized approach was followed to retrieve and analyze focused research. The data for analysis was extracted from the Scopus database and was analyzed using Biblioshiny and VOSViewer. Findings The study found that the countries such as the UK, Australia, China and the USA have special relevance in terms of the number of citations and collaboration networks. Cryptocurrency/Cryptocurrencies, bitcoin have been the base themes along with other crucial issues such as volatility, hedging, COVID-19 pandemic, Ethereum, blockchain, co-integration, portfolio diversification/optimization, spillover, safe haven, investor attention, gold, etc. There is a lot of interdisciplinary research on the theme. Originality/value The current study used a concentrated approach to study the bibliometric literature about the financial implications of cryptocurrency as an asset class and not prominently its technological or legal aspects.
Michael Demmler, Universidad Autónoma de Querétaro-Facultad de Contaduría y Administración
This study explores the financial performance of cryptocurrencies during the COVID-19 pandemic.In particular, the research objective is to compare the market price movements of the leading cryptocurrencies Bitcoin, Ethereum, BNB and XRP before and during the COVID-19 pandemic based on a longitudinal, exploratory, and quantitative research design which is centered on the analysis of the statistical moments of logarithmic return distributions, tests for structural changes combined with stationarity tests and portfolio optimization strategies.Results of the analysis show a clear change of the medium-to long-term return behavior of the analyzed cryptocurrencies during the pandemic, although not immediately after the pandemic announcement of the WHO in March 2020.Especially Bitcoin, BNB and Ethereum show comparable and even more favorable return characteristics in most samples compared to traditional investment alternatives.Furthermore, the diversification potential of cryptocurrency portfolios appears to be quite limited.
Luis Miguel Jiménez Gómez, Erick Lambis-Alandete, Juan D. Velásquez-Henao
Debido al alto atractivo de las criptomonedas, los inversionistas y los investigadores han prestado mayor atención en la previsión de los precios de las criptomonedas. Con el desarrollo metodológico del Deep Learning, la previsión de las criptomonedas ha tenido mayor importancia en los últimos años. En este artículo, se evalúan cuatro modelos de Deep Learning: RNN, LSTM, GRU y CNN-LSTM con el objetivo de evaluar el desempeño en el pronóstico del precio de cierre diario de las dos criptomonedas más importantes: Bitcoin y Ethereum. Se utilizaron métricas de análisis de desempeño como MAE, RMSE, MSE y MAPE y como métrica de ajuste, el R2. Cada modelo de Deep Learning fue optimizado a partir de un conjunto de hiperparámetros y para diferentes ventanas de tiempo. Los resultados experimentales mostraron que el algoritmo RNN tuve un rendimiento superior en la predicción del precio de Bitcoin y el algoritmo LSTM en el precio de Ethereum. Incluso, ambos métodos presentaron mejor desempeño con dos modelos de la literatura evaluados. Finalmente, la confiabilidad del pronóstico de cada modelo se evaluó analizando la autocorrelación de los errores y se encontró que los dos modelos más eficientes tienen alto poder de generalización.
Bu çalışmanın amacı COVID-19 pandemisi döneminde yatırımcı kararlarında meydana gelen değişimleri pandemi öncesi ve sonrası dönemler şeklinde ortaya koyarak finansal sistem içerisinde yer alan ve etkilenen tarafların kararlarında yol gösterici veriler ortaya koymak ve literatüre katkıda bulunmaktır. Çalışma Türkiye örneği üzerinden COVID-19 pandemisi öncesi ve sonrasını içerecek şekilde ve en son güncel değerlerle 01/01/2018-24/02/2023 dönemini kapsamaktadır. Analizler Toda-Yamamoto prosedürünü Fourier fonksiyonu (FTY) ile zenginleştiren bir nedensellik testi kullanılarak yapılmıştır. Çalışma yapılan dönem Chow yapısal kırılma testi ile dört alt döneme ayrılmıştır. Çalışmada USD, Altın (AU) ve Bitcoin değişkenleri ile BIST 100 endeksi arasındaki nedensellik ilişkisi analiz edilmiştir. Yapılan analiz sonuçları pandemi öncesi ve sonrası dönemin birbirinden oldukça farklı nedensellik ilişkileri ortaya koyduğunu, pandeminin ilk şok dalgasında altının güvenli liman özelliğinin ortaya çıktığını, devam eden pandemi sürecinde ise altının bu özelliğini kaybettiği ve ele alınan tüm değişkenler arasındaki nedenselliklerin belirginleştiği görülmüştür. Pandemi sonrası dönemde ise pandemi öncesi döneme kıyasla sadece altının aynı şekilde tek taraflı nedensellik ilişkisine sahip olduğu diğer değişkenler olan USD ve Bitcoin’in BIST100 değişkeniyle nedensellik ilişkisinin tamamen kaybolduğu görülmüştür. Çalışma kriz dönemlerinin her bir aşamasında yatırımcı davranışlarının analiz edilmesi açısından literatüre önemli bir katkı sunmaktadır.
Purpose While an increasing number of investors value socially responsible investment practices, Bitcoin has faced criticism for its carbon footprint resulting from excessive mining power consumption. By examining Bitcoin’s interconnectedness with environmental, social and governance (ESG) equities, this study aims to construct a socially responsible investment strategy for cypto investors. Design/methodology/approach This study uses wavelet analysis and a time-varying parameter vector autoregressive (TVP-VAR) model to uncover the interdependence between ESG equities and Bitcoin. This study computes the optimal ratio, showing that Bitcoin significantly reduces portfolio risk when combined with green stocks. Findings The results show that co-movements between green stocks and Bitcoin are low, indicating that they are suitable combinations for portfolio diversification. From an environmental perspective, this investment strategy offers a theoretical solution to mitigate the negative impacts associated with Bitcoin mining. It aims to address the dilemma faced by sustainability-conscious investors, who must navigate the economic payoff of Bitcoin against their commitment to green investment principles. Practical implications The findings can provide valuable insights for policymakers seeking to develop strategies that promote sustainable investments among crypto investors. Originality/value Research on ethical investment practices in the cryptocurrency market remains in the early stages of development. Ethical investors can benefit from including Bitcoin in their ESG equity portfolios.
Sitara Karim, Brian M. Lucey, Muhammad Abubakr Naeem, Larisa Yarovaya
Abstract The current study investigates the extreme risk dependence between green bonds and financial markets by employing the dual approaches of time‐varying optimal copula and extreme risk spillover analysis of dynamic conditional Value‐at‐Risk. We report significant symmetric (asymmetric) tail‐dependent copulas in the upper (lower) tails characterizing independent regimes. Green bonds offer sufficient diversification, safe‐haven, and hedging opportunities during stable and distressing times to financial markets. The extreme risk spillovers revealed that COVID‐19 transformed the spillovers between green bonds and financial markets except Bitcoin. We proposed insightful implications for policymakers, governments, investors, and portfolio managers to relish the findings for their investment avenues.
In this study, it was investigated which of the machine learning algorithms gave the most optimal results in determining the effect of news about cryptocurrencies on the prediction of Bitcoin (BTC) price. In this direction, three different data sets were created, and these data were tested separately in each machine learning algorithm examined. In this way, it is aimed to determine the effect of the data set on the success of machine learning algorithms. By doing web scraping with the script, we developed to determine the news data we used in the study, 684 news made by BBC, CNBC, Fox, Guardian, NY Post, Over News, Sun, USA Today, and VOA news sources were found and by sentiment analysis of them, 564 news related to cryptocurrencies were taken into account in the study. March 2020 and March 2021 were chosen as the dates of the selected news. (The period when the BTC price rose from 9Ks to 30Ks.). The reason for choosing a past time period is to precisely measure the consistency of the predictions made. While creating the data sets used in machine learning algorithms, high, low, volume, open, and close values in the dollar, Ethereum, and gold values were also taken into account, as well as news sources. In this way, it is aimed to make more realistic estimations by taking market data into account. Machine learning methods in the study; Decision tree, deep learning, random forest, support vector machine, gradient boosted trees, and a generalized linear model was used. The generalized linear model (model II), which gives the most successful result with a rate of 98.73%, is explained in detail in the study.
In this paper, we use quantile-based analysis and nonlinear conditional autoregressive value-at-risk (CAViaR) model to investigate how cryptocurrency price and policy uncertainty indexes affect exchange traded funds (ETFs) investing in equity securities of companies operating in various disruptive innovation sectors and related to digital assets and blockchain technology. Results of the quantile-on-quantile (QQ) estimates and quantile regression reveal that cryptocurrency price and policy uncertainty generally have a negative impact on the globally traded ETFs suggesting that the changes in cryptocurrency uncertainty lead to destabilizing effects in the ETF market. Moreover, we find that the impact of cryptocurrency uncertainty is asymmetric depending on each quantile of ETF returns even though each ETF is influenced in the same way by each uncertainty index, indicating that cryptocurrency price uncertainty and cryptocurrency policy uncertainty have common effects on the ETF market. A linear quantile causality analysis indicated that there is evidence of a significant causality running from the considered uncertainty indexes to the ETF returns, revealing that cryptocurrency price and policy uncertainty have a significant information value in explaining expected ETF returns. On the other hand, we find that linear quantile models outperform nonlinear CAViaR models in quantiles that are only moderate.