This study presents an innovative approach for predicting cryptocurrency time series, specifically focusing on Bitcoin, Ethereum, and Litecoin. The methodology integrates the use of technical indicators, a Performer neural network, and BiLSTM (Bidirectional Long Short-Term Memory) to capture temporal dynamics and extract significant features from raw cryptocurrency data. The application of technical indicators, such facilitates the extraction of intricate patterns, momentum, volatility, and trends. The Performer neural network, employing Fast Attention Via positive Orthogonal Random features (FAVOR+), has demonstrated superior computational efficiency and scalability compared to the traditional Multi-head attention mechanism in Transformer models. Additionally, the integration of BiLSTM in the feedforward network enhances the model's capacity to capture temporal dynamics in the data, processing it in both forward and backward directions. This is particularly advantageous for time series data where past and future data points can influence the current state. The proposed method has been applied to the hourly and daily timeframes of the major cryptocurrencies and its performance has been benchmarked against other methods documented in the literature. The results underscore the potential of the proposed method to outperform existing models, marking a significant progression in the field of cryptocurrency price prediction.
Since 2012, researchers have explored various factors influencing Bitcoin prices. Up until the end of July 2023, more than 9100 research papers on cryptocurrencies were published and indexed in the Web of Science Clarivate platform. The objective of this paper is to analyze the impact of publications on Bitcoin prices. This study aims to uncover significant themes within these research articles, focusing on cryptocurrencies in general and Bitcoin specifically. The research employs latent Dirichlet allocation to identify key topics from the unstructured abstracts. To determine the optimal number of topics, perplexity and topic coherence metrics are calculated. Additionally, the abstracts are processed using BERT-transformers and Word2Vec and their potential to predict Bitcoin prices is assessed. Based on the results, while the research helps in understanding cryptocurrencies, the potential of academic publications to influence Bitcoin prices is not significant, demonstrating a weak connection. In other words, the movements of Bitcoin prices are not influenced by the scientific writing in this specific field. The primary topics emerging from the analysis are the blockchain, market dynamics, transactions, pricing trends, network security, and the mining process. These findings suggest that future research should pay closer attention to issues like the energy demands and environmental impacts of mining, anti-money laundering measures, and behavioral aspects related to cryptocurrencies.
The study investigates the predictability of both the individual and basket of 10 major cryptocurrencies’ daily price changes between 2017 and 2023 by employing various machine learning classification algorithms such as random forests, k-nearest neighbour, decision trees, logistic regression, and Bernoulli naïve Bayes. These models utilize 15 different features based on historical price data and technical indicators as input features. The study estimates find logistic regression as superior over other models under consideration in predicting cryptocurrency daily returns. Overall, the study finds that on an average machine learning classification algorithms predictive accuracies have surpassed 50% when applied to daily frequencies on the basket of 10 major cryptocurrencies.
This research aims to examine the influence of cryptocurrencies on stock market prices in Indonesia. This research uses multiple regression analysis using daily tme series data from 2020-2022 so that the number of observations in this research is 1096. The findings in this research show mixed results between cryptocurrency assets and stock market prices in Indonesia. From the research results, Bitcoin does not have a significant influence on stock market prices in Indonesia, while Ethereum and Binance Coin have a positive and significant influence, but this is different from Maker and Pax Gold. Maker and Pax Gold have a negative and significant influence on stock market prices. The findings in this research show that the nature of the influence of cryptocurrency on stock market prices in Indonesia is not the same but depends on the cryptocurrency asset itself. The findings in this research suggest that investors and market players need to consider these two assets together in their investment strategies.
Abstract Compares ARMA models, boosting, neural network models, HAR_RV models and proposes a new method for predicting one day ahead realized volatility of financial series. HAR_RV models are taken as compared classical volatility prediction models. In addition, the phenomenon of transfer learning for boosting and neural network models is investigated. Bitcoin and E-mini S&P500 are chosen as examples. The realized volatility is calculated based on intraday (intraday—24 hours) data. The calculation is based on the closing values of the internal five-minute intervals. Comparisons are made both within and between the two intervals. The intervals considered are January 1, 2018–January 1, 2022 and January 1, 2018–April 2, 2023. Since there were structural changes in the markets during these intervals, the models are estimated in sliding windows of 399 days length. For each time series, we compare three-parameter enumeration boosting, about 10 different neural network architectures, ARMA models, the newly proposed CTCM method, and various training transfer and training sample expansion options. It is shown that ARMA and HAR_RV models are generally inferior to other listed methods and models. The CTCM model and neural networks of CNN architecture are the most suitable for financial time series forecasting and show the best results. Although transfer learning shows no improvement in terms of forecast precision and yields little decline. It requires more extensive and detailed study. The smallest MAPEs for Bitcoin and E-mini S&P500 realized volatility forecasts are achieved by the newly proposed CTCM model and are 21.075%, 25.311% on the first interval and 21.996%, 26.549% on the second interval, respectively.
Over time, cryptocurrencies like Bitcoin have attracted investor's and speculators' interest. Bitcoin's dramatic rise in value in recent years has caught the attention of many who see it as a promising investment asset. After all, Bitcoin investment is inseparable from Bitcoin price volatility that investors must mitigate. This research aims to use Lee's Fuzzy Time Series approach to forecast the price of Bitcoin. A time series analysis method called Lee's Fuzzy Time Series to get around ambiguity and uncertainty in time series data. Ching-Cheng Lee first introduced this approach in his research on time series prediction. This method is a development of several previous fuzzy time series (FTS) models, namely Song and Chissom and Cheng and Chen. According to most previous studies, Lee's model was stated to be able to convey more precise forecasting results than the classic model from the FTS. This study used first and second orders, where researchers obtained error values from the first order of 5.419% and the second order of 4.042%, which means that the forecasting results are excellent. But of both orders, only the first order can be used to predict the next period's Bitcoin price. In the second order, the resulting relations in the next period do not have groups in their fuzzy logical relationship group (FLRG), so they can not predict the price in the next period. This study contributes to considering investors and the general public as a factor in keeping, selling, or purchasing cryptocurrencies.
The accurate imputation of missing values in time series data is paramount for maintaining the integrity and reliability of analyses and predictions. This article investigates the effica-cy of various missing values imputation methods, encom-passing well-known machine learning and statistical tech-niques. Moreover, for a better understanding, they imple-mented two financial data time series: S&P 500 and Bitcoin markets spanning from 2016 to 2023 on a daily frequency. Initially utilizing complete datasets, controlled missingness was introduced by randomly removing 45 data points. Then, these methods applied multiple imputation strategies for estimating and substituting these missing values. Experi-mental evaluation yielded insightful findings regarding the performance of the different methods. The examined ma-chine learning methods, including k-Nearest Neighbors (k-NN), Random Forest, Deep Learning, and Decision Trees, consistently outperformed their statistical counterparts, such as Mean Imputation, Regression Imputation, Hot-Deck Im-putation, and Expectation-Maximization Imputation. Nota-bly, Random Forest emerged as the most effective method, showcasing superior performance in terms of accuracy and robustness. Conversely, the Mean Imputation method exhibited com-paratively inferior outcomes, suggesting its limited suitabil-ity for financial time series data. This research contributes to the ongoing discourse on data integrity within finance ana-lytics and serves as a comprehensive guide for practitioners seeking optimal missing values imputation methods. The empirical evidence provided herein advances the under-standing of imputation techniques' relative performance and their application in financial data, facilitating enhanced de-cision-making processes and yielding more reliable predic-tions.
Taicir Mezghani, Mustafa Raza Rabbani, Yousra Trichilli, Boujelbène Abbes
The study investigates the dynamic interconnections and opportunities for hedging among cryptocurrency, commodity, and Islamic stock markets using DCC-GARCH and Spillover connectedness models. Using daily data covering the Russia-Ukraine war and COVID-19 outbreak from December 1, 2019 to April 15, 2022, we document weak and frequently negative correlation between Bitcoin and Islamic stock markets. Thus, Bitcoin could be viewed as a haven from Islamic stock market losses. The results also indicate that Bitcoin's diversification benefits are normally steady and increase considerably during turbulence. Furthermore, the net return spillovers from the Bitcoin market remain above zero during most of the study period. We also find that utilizing Bitcoin as a hedge during the COVID-19 pandemic phase leads to higher expenses. The outcomes of this investigation are expected to carry substantial ramifications for Indonesian investors and portfolio managers who adhere to Shariah law since they will enable them to comprehend the advantages of diversifying portfolios across various periods of stock holding or investment horizons.
<title>Abstract</title> The emergence of Bitcoin as a decentralized digital currency has underscored the importance of developing advanced techniques for predicting its price fluctuations. This study evaluates the predictive power of Bitcoin-related Google search volumes and Twitter sentiment analysis within short time frames. By leveraging machine learning algorithms and opinion mining, we identify correlations between online behaviors and Bitcoin price movements. Our methodology encompasses data sourcing, preprocessing, exploratory analysis, feature selection using Correlation Analysis, F-regression, Shapley values, and price prediction with a Long Short-Term Memory (LSTM) model. Findings reveal that Google search data, compared to Twitter sentiment, significantly enhances model accuracy and reduces prediction errors. The study suggests future research to investigate other search engines and online news sentiment, acknowledging limitations in data quality and accessibility of historical Twitter data.
Abstract Predictive demand forecasting plays a pivotal role in optimizing supply chain management, enabling businesses to effectively allocate resources and minimize operational inefficiencies. This paper introduces a novel approach to enhancing demand forecasting processes by leveraging the Ethereum virtual machine within a blockchain framework. The proposed system capitalizes on the inherent security, transparency, and decentralized nature of blockchain technology to create a secure and efficient platform for predictive demand forecasting. The system leverages the Ethereum virtual machine to establish a secure, decentralized, and tamper‐resistant platform for demand prediction while ensuring data integrity and privacy. By utilizing the capabilities of smart contracts and decentralized applications within the Ethereum ecosystem, the proposed system offers an efficient and transparent solution for demand forecasting challenges. The current research focused on Ethereum virtual machine characteristics, features, components, and implementation details. A secured framework for the prediction of demand forecasting systems is proposed. Finally, the authors tried to validate and optimize the gas cost by using distinguished statistics and analysis.
Ali Mansourabady, Fatemeh Tabe, Amir Hossein Rasekh, Ali Ghermezian
Traders and investors are always looking for a way to predict the price of cryptocurrencies to increase their returns and reduce their risks. However, due to unpredictability, instability, and movement, cryptocurrency price prediction is a challenging task. Researchers have proposed different architectures for prediction based on statistical approaches, machine learning (ML), and deep learning (DL) techniques. In this article, we aim to evaluate some of these approaches by implementing their proposed architectures on historical data of Monero (XMR) cryptocurrency from the beginning of 2016 to the end of November 2023 and compare the results. According to the obtained results, the CNN-LSTM-Dense architecture performs better based on the Mean Squared Error (MSE) evaluation metric by achieving a value of 0.00472.
Rodrigo Colnago Contreras, Vitor Trevelin Xavier da Silva, Igor Trevelin Xavier da Silva, Monique Simplicio Viana · 8 authors
Since financial assets on stock exchanges were created, investors have sought to predict their future values. Currently, cryptocurrencies are also seen as assets. Machine learning is increasingly adopted to assist and automate investments. The main objective of this paper is to make daily predictions about the movement direction of financial time series through classification models, financial time series preprocessing methods, and feature selection with genetic algorithms. The target time series are Bitcoin, Ibovespa, and Vale. The methodology of this paper includes the following steps: collecting time series of financial assets; data preprocessing; feature selection with genetic algorithms; and the training and testing of machine learning models. The results were obtained by evaluating the models with the area under the ROC curve metric. For the best prediction models for Bitcoin, Ibovespa, and Vale, values of 0.61, 0.62, and 0.58 were obtained, respectively. In conclusion, the feature selection allowed the improvement of performance in most models, and the input series in the form of percentage variation obtained a good performance, although it was composed of fewer attributes in relation to the other sets tested.
Purpose The present study distinguishes itself by pioneering an innovative framework that integrates key elements of prospect theory and the fundamental principles of electronic word of mouth (EWOM) to forecast Bitcoin/USD price fluctuations using Twitter sentiment analysis. Design/methodology/approach We utilized Twitter data as our primary data source. We meticulously collected a dataset consisting of over 3 million tweets spanning a nine-year period, from 2013 to 2022, covering a total of 3,215 days with an average daily tweet count of 1,000. The tweets were identified by utilizing the “bitcoin” and/or “btc” keywords through the snscrape python library. Diverging from conventional approaches, we introduce four distinct variables, encompassing normalized positive and negative sentiment scores as well as sentiment variance. These refinements markedly enhance sentiment analysis within the sphere of financial risk management. Findings Our findings highlight the substantial impact of negative sentiments in driving Bitcoin price declines, in contrast to the role of positive sentiments in facilitating price upswings. These results underscore the critical importance of continuous, real-time monitoring of negative sentiment shifts within the cryptocurrency market. Practical implications Our study holds substantial significance for both risk managers and investors, providing a crucial tool for well-informed decision-making in the cryptocurrency market. The implications drawn from our study hold notable relevance for financial risk management. Originality/value We present an innovative framework combining prospect theory and core principles of EWOM to predict Bitcoin price fluctuations through analysis of Twitter sentiment. Unlike conventional methods, we incorporate distinct positive and negative sentiment scores instead of relying solely on a single compound score. Notably, our pioneering sentiment analysis framework dissects sentiment into separate positive and negative components, advancing our comprehension of market sentiment dynamics. Furthermore, it equips financial institutions and investors with a more detailed and actionable insight into the risks associated not only with Bitcoin but also with other assets influenced by sentiment-driven market dynamics.
Indri Tri Julianto, Dede Kurniadi, Ricky Rohmanto, Fathia Alisha Fauzia
Ethereum is a cryptocurrency that is now the second most popular digital asset after Bitcoin. High trading volume is the trigger for the popularity of this cryptocurrency. In addition, Ethereum is home to various decentralized applications and acts as a link for Decentralized Finance (DeFi) transactions, Non-Fungible Tokens (NFTs) and the use of smart contracts in the crypto space. This study aims to improve the performance of the forecasting algorithm by using feature extraction for Ethereum price forecasting. The algorithms used are neural networks, deep learning, and support vector machines. The research methodology used is Knowledge Discovery in Databases. The data set used comes from the yahoo.finance.com website regarding Ethereum prices. The results show that the neural network Algorithm is the best Algorithm compared to Deep Learning and support vector machine. The root mean square error value for the neural network before feature selection is 93,248 +/- 168,135 (micro average: 186,580 +/- 0,000) Linear Sampling method and 54,451 +/- 26,771 (micro average: 60,318 +/- 0,000) Shuffled Sampling method. Then after feature selection, the root mean square error value improved to 38,102 +/- 31,093 (micro average: 48,600 +/- 0,000) using the Shuffled Sampling method
The popularity of cryptocurrencies has skyrocketed in the last several years due to the introduction of blockchain technology (BCT). Herein, we are navigating the intersection of sustainable market investment and cryptocurrency predictive analysis against the backdrop of a dynamic and evolving financial landscape marked by the surge of digital assets. This study's goal is to construct the predictive analysis model (PAM) which incorporates Long Short-Term Memory (LSTM) capabilities to predict the price of Bitcoin with high accuracy the next day and to identify the variables that influence price. In constructed PAM, we are using a comprehensive methodology to study temporal correlations within minute-by-minute bitcoin data using preprocessing, sophisticated machine learning algorithms, and data exploration. Our findings demonstrate the effectiveness of the LSTM model in forecasting bitcoin behavior, offering detailed information that is essential for long-term market investing.
The rapid development of information technology, especially the Internet, has facilitated users with a quick and easy way to seek information. With these convenience offered by internet services, many individuals who initially invested in gold and precious metals are now shifting into digital investments in form of cryptocurrencies. However, investments in crypto coins are filled with uncertainties and fluctuation in daily basis. This risk posed as significant challenges for coin investors that could result in substantial investment losses. The uncertainty of the value of these crypto coins is a critical issue in the field of coin investment. Forecasting, is one of the methods used to predict the future value of these crypto coins. By utilizing the models of Long Short Term Memory, Support Vector Machine, and Polynomial Regression algorithm for forecasting, a performance comparison is conducted to determine which algorithm model is most suitable for predicting crypto currency prices. The mean square error is employed as a benchmark for the comparison. By applying those three constructed algorithm models, the Support Vector Machine uses a linear kernel to produce the smallest mean square error compared to the Long Short Term Memory and Polynomial Regression algorithm models, with a mean square error value of 0.02. Keywords: Cryptocurrency, Forecasting, Long Short Term Memory, Mean Square Error, Polynomial Regression, Support Vector Machine
Econometrics can be used to understand and forecast price movements, assess market efficiency, and explore the factors influencing Bitcoin's value and adaptation. Econometrics is related to bitcoin in seven categories: price analysis and prediction, market efficiency, determination of Bitcoin prices, risk analysis, adaptation and network effects, causality tests, and simulation and stress. Testing these analyses can be invaluable for policymakers, investors, and financial institutions interested in the economics of digital currencies. Bitcoin price prediction in machine learning has many challenges that have deep roots in 2 main properties: cryptocurrencies and complexities in the Machine Learning models. Many problems are associated with machine learning for bitcoin price prediction, such as overfitting, data quality and availability, latent variables, model interpretability, computational complexity, dynamic adaptation, market manipulation, anomalies, data snooping bias risk, and time horizon mismatch. In the paper, we proposed an efficient framework for the prediction of bitcoin using nine different machine learning algorithms (linear Regression, random forest, adaboost, tree, KNN, gradient boosting, constant, neural network, SVM) on five different datasets. The results revealed that linear Regression emerged as the optimal model for the first data set. In the second data set, the random forest model demonstrated superior performance. The third data set exhibited the highest efficacy when the Adaboost model was employed. The fourth data set yielded the best outcomes with the random forest model, while linear Regression was the most effective choice for the final data set.
The significant increase in cryptocurrency trading on digital blockchain platforms has led to a growing interest in employing machine learning techniques for the effective prediction of highly nonlinear and nonstationary data, becoming increasingly popular among both individual and institutional market participants. The aim of this research is to deal with the challenging task of predicting the closing prices of two prominent cryptocurrencies, Binance Coin (BNB) and Ethereum (ETH), utilizing machine-learning techniques. This study evaluates the efficacy of various machine learning models in predicting cryptocurrency prices, with a particular focus on Support Vector Machines for Regression (SVR), least-squares Boosting (LSBoost), and Artificial Neural Networks and Adaptive Neuro-Fuzzy Inference System (ANFIS). These models are compared under various metrics. ANFIS models exhibited superior predictive performance on both training and testing datasets based on diverse performance metrics. Comparatively, SVR with a linear kernel demonstrated strong generalization capabilities, particularly on the testing set. LSBoost, while showing promise in training accuracy, indicated results with higher test errors. ANN models maintained a balance between training and testing. This comparison showed the models’ effectiveness, particularly the robustness of ANFIS in capturing the volatile cryptocurrency market trends. The experimental data suggest that certain of the above models can be utilized to predict the ETH and BNB closing price in real time with promising accuracy and experimentally proven profitability.
Bitcoin’s daily value fluctuations are very dynamic. Understanding its rapid and intricate price movements demands advanced techniques for processing complex data. This research aims to compare the accuracy of two machine learning methods, Random Forest (RF) and Long Short-Term Memory (LSTM), in predicting Bitcoin price. This research employs RF and LSTM algorithms to forecast Bitcoin prices using a two-year Yahoo Finance dataset. The evaluation metrics used were accuracy based on Mean Absolute Percentage Error (MAPE) and computational power (CPU-Z). As a result of this research, the LSTM model demonstrates higher accuracy compared to the RF model. MAPE reveals LSTM’s precision of 99.8% and RF’s accuracy of 90.1%. Regarding computational time and resources, RF shows slightly better performance than LSTM. The visual comparison further emphasizes LSTM’s better performance in predicting Bitcoin prices, highlighting its potential for informed decision-making in cryptocurrency trading. This research contributes valuable insights into the effectiveness, strengths, and weaknesses of LSTM and RF models in predicting cryptocurrency trends.
As Deep Learning (DL) continues to be widely adopted, the growing field of study on the robustness of DL approaches in finance is gaining steam. This paper investigates the robustness of a Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) intended for daily closing price predictions of Bitcoin (BTC). The research entails reproducing and adjusting an LSTM design from previous research, with an emphasis on evaluating the robustness of the network. The network is trained using data that has been disturbed by Gaussian noise to assess robustness, and the effect on predictions made outside of the sample is examined. To examine the impact of adding Gaussian noise layers and noisy dense layers on training accuracy and out-of-sample predictions, further robustness tests are conducted. The results show that the LSTM network has remarkable robustness to random disturbances in the data. Nevertheless, the Root Mean Square Error (RMSE) of the prediction increases with the addition of Gaussian noise and noisy dense layers. When random noise is present in the training data, the Autoregressive Integrated Moving Average (ARIMA) model is more vulnerable to it than the LSTM, according to the robustness of the two models. These findings highlight how robustness DL techniques are overall when compared to more conventional linear methods. However, because these models are black-box, the study highlights the significance of comprehensive testing. Although the robustness of the LSTM is impressive, it is important to understand that each network may behave differently depending on the circumstances.
Cryptocurrencies have become a popular investment option and the Ethereum has become a mainstream cryptocurrency because of the additional functionality that can be accomplished with the backing of the powerful Ethereum network compared to Bitcoin. The high volatility of Ethereum offers both profits and risks, making it crucial to improve the forecasting ability for its price volatility. The results of this study could be useful for investors and policymakers who are interested in understanding and managing the risks associated with investing in Ethereum. Several studies have explored similar topics using heterogeneous autoregressive (HAR) models for cryptocurrencies, but this paper offers a more expansive approach. This paper employs five-minute high-frequency data to construct 4 HAR models to predict the volatility of Ethereum, taking into account the impact of structural breaks, Bitcoin, SP500 and VIX. The model that considers all factors outperforms other models for out-of-sample predictions for the 1-week forecasting. Due to the nature of the Ethereum price, the HAR-RV model has achieved a perfect fit in 1-day and 1-month forecasting. Therefore, other models have a very small improvement in fitness and prediction accuracy. This paper contributes to the understanding of Ethereum’s volatility and its impact on the cryptocurrency market.
M. Shyamala Devi, J. Arun Pandian, D. Umanandhini, B. Surekha · 5 authors
Bitcoin is the very first decentralized digital money in the entire world. Bitcoin is exceptionally reliable since it uses a block chain instead of a service provider like a banking institution. As a type of digital precious metal like gold, the price of a single bitcoin seems to have been significantly increasing since the year 2010. As a result, bitcoin is extremely risky for speculators as its value varies often. Since prices may now be predicted in real time, traditional techniques to price forecasting have failed to provide sufficient information and responses. This paper recommends Deep Q-Network based Reinforcement Learning technique (DQNRL) to predict the bitcoin price more effectively. The Binance Bitcoin Futures Price Dataset from Kaggle that contains 3082284 samples and is used towards the bitcoin future cost prediction for DQNRL. The DQNRL model starts with the data exploration that portray the bitcoin price with the standard deviation and mean, log transformation of data. After data preparation, the Binance Bitcoin Futures Price dataset is fitted with the existing regression technique and proposed DQNRL model and the performance is analyzed with MAE, MSE and RMSE. The proposed DQNRL exhibits low MAE with 0.2471, low MSE with 0.061 and low RMSE of 0.2469 when compared to other existing regression models after implementation.