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.
Kriptovalute su posljednja novost u razvoju novca koga ne emitira i regulira nijedna središnja banka ni institucija te nemaju pokriće u zlatu ili srebru već se zasnivaju na matematici i algoritmima, a smatra se rizičnim ulaganjem čija cijena ovisi o dobroj tehnologiji i sigurnosti koja stoji iza te kriptovalute. Ulagače u kriptovalute zanima je moguće na temelju povijesnih podataka kretanja tečaja kriptovalute predvidjeti cijenu u budućnosti pri čemu se koriste razne tehnike uključujući LSTM algoritam koji je korišten u ovom projektu. Za izradu projekta korišten je dataset koji sadrži povijesne podatke cijene Bitcoina u američkim dolarima od 2010. do 2019. godine, a ima za cilj stvoriti model koji će predvidjeti cijenu Bitcoina u USD-u na temelju tih povijesnih podataka. Da bi se utvrdila preciznost predviđanja modela uspoređujemo predviđene cijene s stvarnim cijenama iz dataseta
We develop a general and practical framework to address the problem of the optimal design of dynamic fee mechanisms for multiple blockchain resources. Our framework allows to compute policies that optimally trade-off between adjusting resource prices to handle persistent demand shifts versus being robust to local noise in the observed block demand. In the general case with more than one resource, our optimal policies correctly handle cross-effects (complementarity and substitutability) in resource demands. We also show how these cross-effects can be used to inform resource design, i.e. combining resources into bundles that have low demand-side cross-effects can yield simpler and more efficient price-update rules. Our framework is also practical, we demonstrate how it can be used to refine or inform the design of heuristic fee update rules such as EIP-1559 or EIP-4844 with two case studies. We then estimate a uni-dimensional version of our model using real market data from the Ethereum blockchain and empirically compare the performance of our optimal policies to EIP-1559.
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.
Yeajun Kang, Wonwoong Kim, Hyunji Kim, Minwoo Lee · 6 authors
A smart contract is a digital contract on a blockchain. Through smart contracts, transactions between parties are possible without a third party on the blockchain network. However, there are malicious contracts, such as greedy contracts, which can cause enormous damage to users and blockchain networks. Therefore, countermeasures against this problem are required. In this work, we propose a greedy contract detection system based on deep learning. The detection model is trained through the frequency of opcodes in the smart contract. Additionally, we implement Gredeeptector, a lightweight model for deployment on the IoT. We identify important instructions for detection through explainable artificial intelligence (XAI). After that, we train the Greedeeptector through only important instructions. Therefore, Greedeeptector is a computationally and memory-efficient detection model for the IoT. Through our approach, we achieve a high detection accuracy of 92.3%. In addition, the file size of the lightweight model is reduced by 41.5% compared to the base model and there is little loss of accuracy.
Leonardo H.S. Fernandes, JOSÉ W. L. SILVA, Aurelio F. Bariviera, Kleber E S Sobrinho · 5 authors
This paper sheds light on the changes suffered in cryptocurrencies due to the COVID-19 shock through a non-linear cross-correlations and similarity perspective. We have collected daily price and volume data for the seven largest cryptocurrencies considering trade volume and market capitalization. For both attributes (price and volume), we calculate their volatility and compute the Multifractal Detrended Cross-Correlations (MF-DCCA) to estimate the complexity parameters that describe the degree of multifractality of the underlying process. We detect (before and during COVID-19) a standard multifractal behaviour for these volatility time series pairs and an overall persistent long-term correlation. However, multifractality for price volatility time series pairs displays more persistent behaviour than the volume volatility time series pairs. From a financial perspective, it reveals that the volatility time series pairs for the price are marked by an increase in the non-linear cross-correlations excluding the pair Bitcoin vs Dogecoin (í µí»¼ í µí±¥í µí±¦ (0) = −1.14%). At the same time, all volatility time series pairs considering the volume attribute are marked by a decrease in the non-linear cross-correlations. The K-means technique indicates that these volatility time series for the price attribute were resilient to the shock of COVID-19. While for these volatility time series for the volume attribute, we find that the COVID-19 shock drove changes in cryptocurrency groups.
Given the volatile nature of cryptocurrencies, accurately forecasting cryptocurrency volatility and understanding its determinants are crucial. This paper applies machine learning (ML) techniques to forecast cryptocurrency volatility using internal determinants (e.g., lagged volatility, previous trading information) and external determinants (e.g., technology, financial, and policy uncertainty factors). Both Random Forest and Long Short-Term Memory (LSTM) networks significantly outperform traditional volatility models such as GARCH. Furthermore, we explore two optimization models—Genetic Algorithm and Artificial Bee Colony—to tune the hyper-parameters of LSTM. Our results indicate that the application of these optimization models substantially improves forecasting performance. Moreover, using SHapley Additive exPlanations, an interpretation method, we find that internal determinants play the most important roles in volatility forecasts. Finally, our results show that models trained with determinants from multiple cryptocurrencies outperform those trained with determinants from a single cryptocurrency, suggesting that considering a broader range of determinants can capture the complex dynamics in the cryptocurrency market.
For more than a decade, as the number and value of cryptocurrencies exploded, more and more investors flocked to the cryptocurrency market with the expectation of positive returns. The price of cryptocurrencies, on the other hand, is extremely volatile. As a result, there is a great need to develop an accurate price prediction model to assist investors in making decisions and profit. This paper focuses on developing an LSTM-based prediction model for Bitcoin, Ethereum, EOS, and Solana cryptocurrency price prediction and calculating their RMSE and MAPE. Furthermore, four models are compared using this calculated MAPE. Based on the comparison results, the impact of cryptocurrency volatility, liquidity, and technology level on the accuracy of the LSTM prediction model is also examined. The paper concludes that the LSTM model can predict the price of Bitcoin more accurately because Bitcoin has the least volatility, the most liquidity and uses the oldest but most secure consensus mechanism.
This article reveals a specific category of solutions for the 1+1 variable order (VO) nonlinear fractional Fokker-Planck equations. These solutions are formulated using VO q-Gaussian functions, granting them significant versatility in their application to various real-world systems, such as financial economy areas spanning from conventional stock markets to cryptocurrencies. The VO q-Gaussian functions provide a more robust expression for the distribution function of price returns in real-world systems. Additionally, we analyzed the temporal evolution of the anomalous characteristic exponents derived from our study, which are associated with the long-term (power-law) memory in time series data and autocorrelation patterns.
This study demonstrates the significant impact of market sentiment, derived from social media, on the daily price prediction of cryptocurrencies in both bull and bear markets. Through the analysis of approximately 567 thousand tweets related to twelve specific cryptocurrencies, we incorporate the sentiment extracted from these tweets along with daily price data into our prediction models. We test various algorithms, including ordinary least squares regression, long short-term memory network and neural hierarchical interpolation for time series forecasting (NHITS). All models show better performance once the sentiment is incorporated into the training data. Beyond merely assessing prediction error, we scrutinise the model performances in a practical setting by applying them to a basic trading algorithm managing three distinct portfolios: established tokens, emerging tokens, and meme tokens. While NHITS emerged as the top-performing model in terms of prediction error, its ability to generate returns is not as compelling.
Purpose- Forecasting techniques and models are extremely important for people and organizations that are in the right decision making and investment stage. Forecast accuracy enables successful decisions and allows investors to maximize their profits. The development of finance and related technologies in the world and innovative financial instruments have made it interesting for investors. The most popular of these developments is undoubtedly Bitcoin, a product of blockchain technology. The purpose of this study is to predict the future values of Bitcoin. Methodology- In this study, future predictions are made using an LSTM model based on Bitcoin's historical data and indicators of key market forecasters. In this study, 3 different data sets were created by selecting 1 indicator from 4 different indicator types. The 10 Bitcoin data coming after the last value is estimated. Findings- In this study, 3 different data sets were created by selecting an indicator from 4 different indicator groups. These datasets were first trained with the iterative neural network LSTM model and then tested with real values. At the same time, the next 10 bitcoin price values were also predicted in a 15-minute period. Error rates at the end of the model were compared with each other. The 1st dataset, with the most used indicators in the datasets, produced the lowest error rate. Conclusion- The dataset 1, which consists of the most used indicators of the datasets, gave the lowest error rate. According to this result, the rate of reaching realistic values increases as the use of indicators increases. Keywords: LSTM, bitcoin, cryptocurrency, neural network, prediction JEL Codes: C53, C45, G10
Cryptocurrencies are highly anonymous, poorly regulated in many countries, and can issue tokens at nearzero cost using existing platforms. As a result, there is no shortage of fraudulent cryptocurrencies that raise large sums of money through hype, then disappear and do little actual project development. The prevalence of fraudulent cryptocurrencies not only harms investors but can also prevent sound companies from raising funds. To remedy this situation, it would be useful to develop a method to determine whether a particular cryptocurrency is fraudulent or not. The information in cryptocurrency whitepapers could be useful in detecting fraudulent cryptocurrency, but there are no clear criteria to evaluate the reliability and feasibility of their content. Besides, most studies analyzing whitepapers focus on the success or failure of ICO ”fundraising” and fail to adequately consider the ongoing development and operation of the project. On the other hand, a few studies have attempted to detect fraudulent cryptocurrencies from whitepapers, but their results suggest the possibility of identifying fraud with high accuracy. The objective of this paper is to build a model to detect fraudulent cryptocurrencies from whitepapers using natural language processing and machine learning techniques, and to verify whether the model has sufficient predictive accuracy in detecting fraud, after solving the problems of previous studies. We collected 250 cryptocurrency whitepapers consisting of 150 frauds and 100 controls, extracted features, and applied multiple machine learning methods to classify frauds and controls. Then analyzed the feature differences between the fraud and control groups, and examined the tendency of fraudulent cryptocurrency whitepapers. We observed 0.841 F1 Score for the best prediction model, which outperforms previous studies. Furthermore, the performance of K-Means, which is unsupervised learning, was not significantly lower than that of other machine learning methods, and a certain level of accuracy was confirmed. Therefore, there is a possibility that K-Means can be used in cases where fraud criteria cannot be clearly defined. We also found that fraudulent cryptocurrency whitepapers used relatively more business and finance-related words. On the other hand, whitepapers in the control group tended to use more blockchain-related technical terms.
In this article, we delve into the challenging problem of forecasting cryptocurrency prices using mathematical extrapolation techniques. We highlight the scarcity of research in this domain, underlining the necessity for in-depth investigation. The article outlines the unresolved issues related to extrapolation-based cryptocurrency price prediction, such as market volatility and non-linearity. It primarily aims to showcase the potential of extrapolation for predicting bitcoin prices. The analysis involves a year-long bitcoin price trend, with the application of linear and polynomial extrapolation methods. While some correlation exists, notable discrepancies, especially during abrupt price changes, are evident. The conclusion emphasizes the limitations of extrapolation and advises a diversified approach to cryptocurrency investment decisions, considering various factors beyond mathematical data. In this article, we delve into the challenging problem of forecasting cryptocurrency prices using mathematical extrapolation techniques. We highlight the scarcity of research in this domain, underlining the necessity for in-depth investigation. The article outlines the unresolved issues related to extrapolation-based cryptocurrency price prediction, such as market volatility and non-linearity. It primarily aims to showcase the potential of extrapolation for predicting bitcoin prices. The analysis involves a year-long bitcoin price trend, with the application of linear and polynomial extrapolation methods. While some correlation exists, notable discrepancies, especially during abrupt price changes, are evident. The conclusion emphasizes the limitations of extrapolation and advises a diversified approach to cryptocurrency investment decisions, considering various factors beyond mathematical data.
Unlike traditional currencies that rely on centralized such as banks or governments, cryptocurrencies have become popular due to its decentralized transactions. Decentralization takes advantage of no requirement for intermediaries, thus reducing transaction fees and processing times. However, investing in cryptocurrencies incurs risks and uncertainties due to price volatility and rapid changes. The fact that prediction of asset prices is complex due to the influence of multiple factors on price movements. This paper studied the technical factor to analyse the short-term returns of Ethereum (ETH) in the periods of 1-10 days. The historical data containing ETH closing price are collected from CoinGecko. The twenty-two indicators are chosen from Momentum, Volatility, and Sentiment factors as candidates to provide valuable insights in market trends. By calculating various indicators based on past closing prices, this study utilizes XGBoost, a powerful boosted decision trees ensemble, to discover patterns in previous trading. The model performance is evaluated using the multi-class AUC-ROC metric, which measures the accuracy of predicting three types of ETH returns: Downtrend, Sideway, and Uptrend. The results show that the models achieve accuracy scores ranging from 0.65 to 0.67. Moreover, the study emphasizes the importance of considering momentum indicators when making investment decisions in Ethereum. Keywords—cryptocurrency investment, technical factor, Ethereum, XGBoost, machine learning
Financial markets are complex, evolving dynamic systems. Due to their irregularity, financial time series forecasting is regarded as a rather challenging task. In recent years, artificial neural network applications in finance for such tasks as pattern recognition, classification, and time series forecasting have dramatically increased. The objective of this paper is to present this versatile framework and attempt to use it to predict the stock return series of four public-listed companies on the New York Stock Exchange. Our findings coincide with those of Burton Malkiel in his book, A Random Walk Down Wall Street; no conclusive evidence is found that our proposed models can predict the stock return series better than that of a random walk.
Time series data from bitcoin has nonlinear data fluctuations so that a model is needed that can accommodate data with these conditions. The method that can be used for nonlinear time series data cases such as bitcoin is the LSTAR-GARCH model. LSTAR-GARCH is a combination of the LSTAR model and the GARCH model. Bitcoin investment also contains an element of risk. To find out the value of risk, the <em>Expected Tail Loss </em>risk measurement tool can be used. <em>Expected Tail Loss </em>(ETL). The data used in this study are historical daily bitcoin price data for the period April 1, 2022 to April 1, 2023. The modeling results obtained based on the MAPE value show that the LSTAR-GARCH model is the best model with the smallest MAPE value of 30% compared to the AR, LSTAR, or AR-GARCH models. The expected Taill loss value of bitcoin is -0.06784.
Abstract In recent years, cryptocurrencies' price prediction has attracted the interest of many people including investors, researchers and practitioners. In this study, we proposed a hybrid model for predicting the daily close price of cryptocurrencies based on different neural networks such as long short‐term memory, convolutional neural network and attention mechanism. Using an ensemble of three pre‐trained language models, we extracted sentiment of cryptocurrency‐related tweets posted between 1 January 2021 and 31 December 2021. We constructed 20 different versions of our model and evaluated their performance on data of 27 most traded cryptocurrencies using a history of previous days' sentiment data along with close prices as input data. The flexible input layer of our model enables different ways of feeding data into the model to adjust it for different cryptocurrencies to obtain better predictions. Our analysis revealed several important findings. We showed that longer sequences of input data achieve most accurate predictions on average. More specifically, using a history of 14‐ and 21‐days' data results in lowest RMSE values on average compared to using a history of 7 days. However, there is no significant difference between the results related to the input sequences with lengths of 14 and 21. In addition, our findings suggest that sentiment data can be useful in predicting prices for more than 70% of the studied cryptocurrencies. Thus, peoples' emotions, opinions, and sentiment that are expressed through their posts on Twitter platform play a significant role in prediction of cryptocurrencies' prices.
OBJECTIVES: With the expansion of social networks such as Twitter, many experts share their opinions on various topics. The opinions of experts, who are also known as influencers, can be very influential. Combining these tweets and the historical prices of cryptocurrencies makes it possible to predict their price trends accurately. A Hybrid of RoBERTa deep neural network and BiGRU has been used for Sentiment Analysis (SA). Sentiments of tweets can be of great help to investors to understand the future behavior of the market and manage the stock portfolio. Unlike the tweets that are only extracted using the cryptocurrency name hashtag, the tweets of this dataset have specialized opinions and can determine the market trend. DATA DESCRIPTION: The dataset created in this research concerns the opinions of more than 52 influencers (persons or companies) regarding eight cryptocurrencies. This dataset was collected through the Apify Twitter API for eight months, from February 2021 to June 2023. This dataset contains five Excel files and tweets, compound score, importance coefficient of each tweet, sentiment polarity, and historical prices of four cryptocurrencies: Bitcoin, Ethereum, Binance, and other information. These tweets cover the opinions of 52 influencers on more than 300 cryptocurrencies, although most comments are related to Bitcoin, Ethereum, and Binance. For this reason, three Excel files containing the historical prices of polarity and compound sentiment related to Bitcoin, Ethereum, and Binance cryptocurrencies have been placed separately in the dataset. The polarity of sentiment in these Excel shows the maximum number of polarities by applying the importance coefficient, which determines the dominant polarity of sentiment related to a particular day for the cryptocurrency.
There has been great interest in developing hierarchical structures of fuzzy rule-based systems due to their flexibility allowing to model complex problems. To cope with the high degree of uncertainty arising from the characteristics of cryptocurrency markets, this paper proposes a hierarchical intuitionistic TSK (Takagi-Sugeno-Kang) fuzzy system equipped with a feature selection and feature ranking component. The proposed system uses intuitionistic fuzzy sets, allowing to effectively model investor uncertainty in the decision-making on cryptocurrency markets. The hierarchical structure is a parallel tree-like fuzzy system that is based on relevant features while considering feature dependencies. Computational efficiency is achieved by using fuzzy c-means clustering to produce rule antecedents. The proposed system is validated using multivariate bitcoin data for the period 2018 to 2022, showing that the proposed system can accurately predict bitcoin prices while retaining an interpretable hierarchical structure.
Cryptocurrencies have become a popular and widely researched topic of interest in recent years for investors and scholars. In order to make informed investment decisions, it is essential to comprehend the factors that impact cryptocurrency prices and to identify risky cryptocurrencies. This paper focuses on analyzing historical data and using artificial intelligence algorithms on on-chain parameters to identify the factors affecting a cryptocurrency's price and to find risky cryptocurrencies. We conducted an analysis of historical cryptocurrencies' on-chain data and measured the correlation between the price and other parameters. In addition, we used clustering and classification in order to get a better understanding of a cryptocurrency and classify it as risky or not. The analysis revealed that a significant proportion of cryptocurrencies (39%) disappeared from the market, while only a small fraction (10%) survived for more than 1000 days. Our analysis revealed a significant negative correlation between cryptocurrency price and maximum and total supply, as well as a weak positive correlation between price and 24-hour trading volume. Moreover, we clustered cryptocurrencies into five distinct groups using their on-chain parameters, which provides investors with a more comprehensive understanding of a cryptocurrency when compared to those clustered with it. Finally, by implementing multiple classifiers to predict whether a cryptocurrency is risky or not, we obtained the best f1-score of 76% using K-Nearest Neighbor.
The popularity of cryptocurrencies has skyrocketed in recent years, with blockchain technologies enabling the development of new digital assets. However, along with their advantages, such as lower transaction costs, increased security, and transactional transparency, cryptocurrencies have also become susceptible to various forms of market manipulation. The pump and dump (P&D) scheme is of significant concern among these manipulation tactics. Despite the growing awareness of P&D activities in cryptocurrency markets, a comprehensive survey is needed to explore the detection methods. This paper aims to fill this gap by reviewing the literature on P&D detection in the cryptocurrency world. This survey provides valuable insights into detecting and classifying P&D schemes in the cryptocurrency market by analyzing the selected studies, including their definitions and the taxonomies of P&D schemes, the methodologies employed, their strengths and weaknesses, and the proposed solutions. Presented here are insights that can guide future research in this field and offer practical approaches to combating P&D manipulations in cryptocurrency trading.
Giacomo di Tollo, Joseph Andria, Gianni Filograsso
Cryptocurrencies are nowadays seen as an investment opportunity, since they show some peculiar features, such as high volatility and diversification properties, that are triggering research interest into investigating their differences with traditional assets. In our paper, we address the problem of predictability of cryptocurrency and stock trends by using data from social online communities and platforms to assess their contribution in terms of predictive power. We extend recent developments in the field by exploiting a combination of stochastic neural networks (NNs), an extension of standard NNs, natural language processing (NLP) to extract sentiment from Twitter, and an external evolutionary algorithm for optimal parameter setting to predict the short-term trend direction. Our results point to good and robust accuracy over time and across different market regimes. Furthermore, we propose to exploit recent advances in sentiment analysis to reassess its role in financial forecasting; in this way, we contribute to the empirical literature by showing that predictions based on sentiment analysis are not found to be significantly different from predictions based on historical data. Nonetheless, compared to stock markets, we find that the accuracy of trend predictions with sentiment analysis is on average much higher for cryptocurrencies.