Louise Gabriel N. De Leon, Rafael C. Gomez, Martin Lance G. Tacal, Jonathan V. Taylar · 6 authors
The cryptocurrency market has shown an imperfect market property due to volatility and erratic behavior. Cryptocurrencies“ prices largely fluctuate because of the 24/7 trading policy that weakens the forecasting power of thetraditional models. However, research works reveal that cryptocurrency prices cannot be truly random in nature. In thispaper, traditional methods of time-series forecasting such as the ARIMA and SARIMA were presented and compared with the recent deep learning models such as the GRU, RNN, LSTM, and Facebook Prophet. Historical data for Bitcoin was acquired consisting of 2770 samples from 2018 to 2022. The study yielded the lowest MSE and RMSE of 2159166.25 and 1469.41 respectively for the GRU model. FB Prophet yielded the highest RMSE and MSE among all the experimented models.
Cryptocurrencies are now the most popular investment instruments among millenials. Crypto offers great returns in a short period of time. Prior to COVID-19, Crypto experienced significant price fluctuations accompanied by an increase in the number of high transaction volumes. This situation was disrupted by the presence of the COVID-19 which made the world economy devastated, marked by the decline of stock prices in the world, especially in Indonesia. A paired test was conducted in this study to compare the state of Crypto before and during COVID-19 with the variables of Risk, Transaction Volume, Return, and Sharpe Performance. The results showed that there was a significant difference in the variables of Transaction Volume and Return. However, there was no significant difference in the Risk and Sharpe performance before and during COVID-19. This study shows that despite the COVID-19 pandemic, the enthusiasm of investors who transact crypto assets is not affected and they still get returns in accordance with the investments made. The high risk will be followed by a high standard deviation, so that the Sharpe Performance is small. Cryptocurrencies still have many gaps to research, such as regulation, so that many countries have not legalized Crypto transactions. If there is no regulation for Crypto, it is certain that an increase in cybercrime harms crypto investors and threatens global financial stability. Nevertheles, with or without COVID-19, investment transactions gain and lose based on confidence in the limited market. Therefore, the success of confidence fluctuations in crypto encourages the emergence of alternative coins created by investors to conduct an Initial Coin Offering (ICO).
In the topic selection, we need to estimate the prices of bitcoin and gold according to the data given from 2001 to 2012. According to the estimated price, the initial amount is set as $1000, which is used as the principal for financial investment for a period of five years from 2016. In the whole modeling process, the main problems we need to solve are the following four points: the task 1 is the best investment strategy is given through the established model, and the investment value on October 9, 2021 is calculated. The task 2 is the best strategy of the model is proved. The task 3 is Determine the impact of transaction costs on transaction results. The last task is to Complete a memo with strategies, models, and results. In the whole modeling process, we first preprocess the data, which is arranged and classified in chronological order, and fill the data by interpolation fitting. LSMT algorithm is a neural network algorithm, which is suitable for the calculation of various long-term processes. The investment problem we study is a good application field. In the calculation process of the basic model, it is necessary to set the initial value, complete a series of processing, and process the hidden layer of LSTM unit. Take x as the output value, set the temporary hidden layer and new hidden layer, and verify that the size of the final output result is consistent with the label size. The hidden layer is transformed according to the sigmoid function proposed above. After calculating the hidden layer conversion, the error is back propagated, the input derivative of the file is obtained, and the overall error and record hidden layer are obtained. Then the calculated hidden layer difference is used to calculate the change of parameters and update parameters. Since bitcoin can be traded on any trading day, gold can only be traded on weekly trading days. For the convenience of calculation, we fix the transactions of bitcoin and gold as trading days every Friday. After receiving the benefits, the total assets of the cash flow as of the trading day are obtained by deducting the Commission to be paid. However, the model ignores the impact of bitcoin mining with different software and the fact that gold and bitcoin are not fixed on the same trading day, so there will be errors.
Cryptocurrency has evolved from a fringe phenomenon to a far more popular method of investing and financing. For investors and traders, predicting the price of bitcoin is critical. Several machine learning algorithms are utilized to anticipate the price of digital money in this research paper. The analysis employed Decision Trees, Light Gradient Boosting Machines, and Neural Networks. The purpose of this study is to look at the predicted accuracy of each machine learning method. According to the analysis, decision tree, lightGBM, and neural networks have a very high accuracy rate when it comes to forecasting cryptocurrencies. These results shed light on guiding further exploration to help investors in building an appropriate digital currency portfolio and reducing risks.
Due to the redefining of money and its price volatility, cryptocurrencies have become one of the most prominent phenomena in recent years. This research investigates how well public opinion on Twitter and news stories may be used to estimate cryptocurrency returns. Three models are designed and compared: LSTM based, LSTM-and-GRU-based, and LSTM and CNN-based models. Firstly, numerals and historical datasets of Bitcoins are used for all three models, which are further extended to Twitter and news datasets. An error score of 1015.17, 1106.71, and 3010.63 is obtained. Then, the proposed models are applied to the combined dataset of Twitter and news from Ethereum, and an error score of 47.85, 34.01, and 58.27 is obtained. Finally, the same methodology is applied to the combined dataset of Litecoin and obtained an error score of 9.45, 8.81, and 15.94. It is observed that LSTM with GRU generates the best results for all the datasets.
To be or not to be is the question that Hamlet thinks about day and night. Gold or Bitcoins is an inescapable choice for investors. With the ever rising and falling price of gold and bitcoin, making good trading decisions is of paramount importance. In this paper, we systematically investigate how data can be used to quantify the factors that influence trading and make the final decision. We build time series with the prices of gold and bitcoin for the past five years. We obtained forecast curves with excellent fit by seasonality analysis and ARIMA time series model forecasts.
Cryptocurrencies are rising in importance as an investment option and alternative currency. Thus, investors are keen on finding timely market movement insights. One such source is Twitter due to its live feed of information on cryptocurrencies and emotional information from investors expressing their sentiments. This article examines the extent to which Twitter sentiments can be used to predict price and transaction volume changes for the nine largest cryptocurrencies for the period of June 2021 to September 2021. This study was conducted using a lexicon-based approach through the VADER algorithm for sentiment analysis, while applying the Granger causality method to analyze the two-way predictive capabilities of each cryptocurrency&s;s sentiments toward their respective price and transaction volume changes. Past studies have shown that sentiment analysis may work for several cryptocurrencies, while this study only found predictive capabilities in transaction volume changes and not in price movement.
Jesper Kristensen, Juan P. Madrigal-Cianci, Giorgos Felekis, Maria Liatsikou
In the scientific and commercial worlds, predicting cryptocurrency prices over time has gotten a lot of interest. Most related studies use variations of Recurrent Neural Networks to forecast the next value of a single coin due to the temporal nature of the challenge. As a result, determining how effectively such a model would perform across various tasks (cryptocurrencies), several future timesteps, and forecasting horizons will be difficult. This paper proposes a multi-task and multi-step sequence-to-sequence model that is trained jointly on 22 cryptocurrencies' time series. Our findings show the value of sequence-to-sequence modeling for future predictions, as well as the significant improvements in accuracy and training time that can be achieved by using a single multi-task model rather than numerous distinct models for each task.
In response to the rise of the bitcoin market, the nonlinear variation of bitcoin price has always been the center of research in the community. Using bitcoin transaction data from 2014 to 2017, this study removes the uncontrollability and unpredictability of external factors and discusses the relationship between the predict and actual price of a single-feature LSTM model and a multi-feature LSTM model that incorporates thermodynamic chart to point out potentially highly correlated variables for the bitcoin price itself only, sets up a one-day prior algorithm, uses Python 3.7, Keras and LSTM tools to plot line plots of predicted and true prices and compare the accuracy of both. We conclude that the LSTM prediction is better with multiple features, which can greatly reduce the error and hedge the risk. Even in the chance case of more drastic fluctuations, the prediction is still better, which improves the utility and applicability of the model.
Babatomiwa Omonayajo, Auwalu Saleh Mubarak, Fadi Al‐Turjman, Zubaida Said Ameen
Ethereum is the most well-known and largest accessible, decentralized, and block chain technology system software, originally introduced in 2015. Unlike Bitcoin, which only permits for monetary exchange, Ethereum enables for the use of Smart Contracts, which are written in the Solidity programming language and executed on the Ethereum blockchain network. When these smart contracts are used to perform a task on the Ethereum blockchain network, the network imposes a transaction price known as gas fees. The prediction of gas prices was conducted in this study using the Facebook prophet model in python programming language with the aim to narrow the gap between the current unexpected gas fee price and the ability of predicting the future gas fee, based on past gas prices and current dataset gathered daily for eight (8) years between 2015 and 2022. The results of this study show a 365-day forecast of the price of ether using three assessment criteria, with results that are near to zero and a fit model.
Digital currencies such as Ethereum and XRP allow for all transactions to be carried out online. To emphasize the decentralized nature of fiat currency, we can refer, for example, to the fact that all virtual currency users may access services without third-party involvement. Cryptocurrency price swings are non-stationary and highly erratic, similarly to the price changes of conventional stocks. Owing to the appeal of cryptocurrencies, both investors and researchers have paid more attention to cryptocurrency price forecasts. With the rise of deep learning, cryptocurrency forecasting has gained great importance. In this study, we present a long short-term memory (LSTM) algorithm that can be used to forecast the values of four types of cryptocurrencies: AMP, Ethereum, Electro-Optical System, and XRP. Mean square error (MSE), root mean square error (RMSE), and normalize root mean square error (NRMSE) analyses were used to evaluate the LSTM model. The findings obtained from these models showed that the LSTM algorithm had superior performance in predicting all forms of cryptocurrencies. Thus, it can be regarded as the most effective algorithm. The LSTM model provided promising and accurate forecasts for all cryptocurrencies. The model was applied to forecast the future closing prices of cryptocurrencies over a period of 180 days. The Pearson correlation metric was applied to assess the correlation between the prediction and target values in the training and testing processes. The LSTM algorithm achieved the highest correlation values in training (R = 96.73%) and in testing (96.09%) in predicting XRP currency prices. Cryptocurrency prices could be accurately predicted using the established LSTM model, which displayed highly efficient performance. The relevance of applying these models is that they may have huge repercussions for the economy by assisting investors and traders in identifying trends in the sales and purchases of different types of cryptocurrencies. The results of the LSTM model were compared with those of existing systems. The results of this study demonstrate that the proposed model showed superior accuracy based on the low prediction errors of the proposed system.
Alfonso Guarino, Luca Grilli, Domenico Santoro, Francesco Messina · 5 authors
Abstract Financial bubbles represent a severe problem for investors. In particular, the cryptocurrency market has witnessed the bursting of different bubbles in the last decade, which in turn have had spillovers on all the markets and real economies of countries. These kinds of markets and their unique characteristics are of great interest to researchers. Generally, investors and financial operators study market trends to understand when bubbles might occur using technical analysis tools. Such tools, which have been historically used, resulted in being precious allies at the basis of more advanced systems. In this regard, different autonomous, adaptive and automated trading agents have been introduced in the literature to study several kinds of markets. Among these, we can distinguish between agents with Zero/Minimal Intelligence (ZI/MI) and Computational Intelligence (CI) -based agents. The first ones typically trade on the market without resorting to complex learning strategies; the second ones usually use (deep) reinforcement learning mechanisms. However, these trading agents have never been tested on the cryptocurrencies market and related financial bubbles, which are still mostly overlooked in the literature. It is unclear how these agents can make profits/losses before, during, and after a bubble to adjust their strategy and avoid critical situations. This paper compares a broad set of trading agents (between ZI/MI and CI ones) and evaluates them with well-known financial indicators (e.g., volatility, returns Sharpe ratio , drawdown, Sortino and Omega ratio ). Among the experiment’s outcomes, ZI/MI agents were more explainable than CI ones. Based on the results obtained above, we introduce GGSMZ , a trading agent relying on a neuro-fuzzy mechanism. The neuro-fuzzy system is able to learn from the trades performed by the agents adopted in the previous stage. GGSMZ ’s performances overcome those of other tested agents. We argue that GGSMZ could be used by investors as a decision support tool.
This paper investigates how to use deep learning methods to combine with traditional multi-factor models and construct a quantitative trading model based on an AutoEncoder algorithm (AE) to classify cryptocurrencies since 2009, so as to screen out ones with investment value and then construct an effective investment portfolio. The AE algorithm is capable of handling high-dimensional data and mining interfactor non-linearities. Our empirical results on cryptocurrencies show that the model outperforms single-type factors and benchmark in terms of Cumulative Returns and the Sharpe Ratio.
Utilizing a large set of variables that include transaction information, public attention, blockchain information, macroeconomic variables and technical indicators, we compare different deep learning models with baseline methods, such as statistical and machine learning models, on Bitcoin volatility forecast. We find that feature selection approach strongly affects model performance. The results show that a simple Long Short-Term Memory (LSTM) model outperforms other models when using individual feature selection method.
Bitcoin is one of the most famous cryptocurrencies in the world and its price has changed frequently in recent years. Different from the traditional stock market, the changes of bitcoin's price are even more dramatic. Huge price movements have attracted not only investors but also more researchers to find different methods to predict bitcoin's price. In this research, we consider using change rate over previous a few days to predict the significant bitcoin price changes, which are defined as a three-category problem. Besides, we compare the predictive results using two deep learning models - LSTM and GRU, and then find LSTM perform better than GRU with the highest AUC value of 0.701 in most cases.
Cryptocurrencies have taken over the whole financial world with their high risk and high reward quality. Cryptocurrency markets are one of the most complex and volatile markets in the world. In this paper, we make an attempt to predict financial time series for popular cryptocurrencies like Ethereum, Binance coin and Bitcoin, using deep learning and multifractal detrended fluctuation analysis techniques. After testing for non-linearity in the financial time series of popular cryptocurrencies, it was revealed that cryptocurrency time series manifest latent relations, short term, and long term memory. The method uses Transformers and Long-short term Neural networks(LSTM) to forecast the prices of various cryptocurrencies. Although using LSTM along with transformers leads to longer computational times, but the predictive accuracy is better as compared to traditional regression neural networks and kNN forecasting models. Experimentation of the proposed model reveals that the deep learning model is profoundly efficient in predicting the intrinsic dynamics of cryptocurrency financial time series.
Jul 20, 2022·Proceedings of the 15th International Conference on Computer Graphics, Visualization, Computer Vision and Image Processing (CGVCVIP 2021), the 7th International Conference on Connected Smart Cities (CSC 2021) and 6th International Conference on Big Data Analytics, Data Mining and Computational Intelligence (BigDaCI’21)
Ningbo Zhu, Fei Yang, Mingzhi Zhu, Xinyao Sun and Irene Cheng
The cryptocurrency industry has evolved rapidly in recent years, and it is increasingly popular as a convenient tool tocomplement the traditional stock and futures exchanges. Accurate market research enables traders to make moreinformed decisions
The cryptocurrency market is generally accepted in the world, and its price has soared and plummeted sharply. Meanwhile, skewness is an index reflecting the rapid rise and fall of asset prices in a short period. Studying the relationship between the skewness of cryptocurrency and its returns can help risk evaders expect bad news and provide a reference for risk enthusiasts to make investment decisions. Therefore, through univariate combination analysis, this paper groups cryptocurrency according to the skewness of the previous month before buying and holding it in the next week. Moreover, the excess return series are calculated to make statistical tests on it. Then we construct a three-factor model of cryptocurrency and adjust the return series. To enhance the robustness of the conclusion, we also use other measures of skewness such as idiosyncratic skewness to conduct a univariate combination analysis. The results show that a positive correlation between cryptocurrency skewness and its returns exists, which can be used as a reference index of the returns.
The Bitcoin (BTC) is one of the most popular cryptocurrencies and now is one type of investment on the stock market. The price prediction is a real challenge as it depends by too many factors, an investment to BTC characterized too risky because the price has too many upside-downs. However BTC was the occasion for many people to start explore the stock market. In the last few years, the prediction of BTC has occupied the scientific community and many approaches have been made. In this research we try to predict the price of BTC per minute with long short-term memory networks, and then pass these predictions to a recurrent reinforcement learning (RRL) model to trade the BTC with United States dollars (USD). In the end of the paper we present the profits that the models made.
Japjeet Singh, Sulalitha Bowala, A. Thavaneswaran, Ruppa K. Thulasiram · 5 authors
The forecasting problems in Computational Finance involve modelling the vagueness and imprecision inherent to the financial markets. Fuzzy set theory has a unique ability to quantitatively and qualitatively model and analyze such problems. Volatility forecasting plays an important role in financial risk management and in option pricing. Recently, there has been a growing interest in data-driven volatility models and neurovolatility models for risk forecasting of stocks and index funds. However, even these state-of-the-art models do not take into account the fuzzy volatility in their risk forecasts.Cryptocurrencies are a novel financial asset class based on the Blockchain technology. Cryptocurrencies have gained popularity among retail investors as a financial asset with high risks and high returns. The extremely volatile nature of cryptocurrencies (compared to traditional assets) makes forecasting their volatility more challenging. A simple algorithmic trading approach, Simple Moving Average (SMA) crossover strategy, is used to calculate the Algo returns. This paper provides fuzzy forecasts of the volatility of Algo returns using the data-driven Exponentially Weighted Moving Average (DD-EWMA) and neuro models for six major cryptocurrencies. We also compute and compare fuzzy volatility forecasts of four major tech stocks and Chicago Board Options Exchange’s (CBOE) volatility index (VIX) using DD-EWMA and neuro models. Our experimental results show that the data-driven models produce better forecasts for cryptocurrencies as compared to the neuro models, while for the regular stocks and indexes, no such definitive conclusion could be drawn.
Sotirios Oikonomopoulos, Katerina Tzafilkou, Dimitrios Karapiperis, Vassilios S. Verykios
In a paper that was anonymously published and signed by the pseudonym Satoshi Nakamoto, Bitcoin was introduced to the world. Due to its enormous success, a great number of cryptocurrencies were created in the upcoming years. This exponential growth relies mostly on the extreme volatility of the market, which led many people to become interested and get involved, primarily for profit. Cryptocurrency enthusiasts tend to share and learn news and opinions on social media platforms, one of the most popular being Twitter. In this paper, we study the extent to which Twitter sentiment analysis can be used to predict price fluctuations for cryptocurrencies. Initially, we gathered tweets and price data of seven of the most popular cryptocurrencies, which were processed to perform sentiment analysis using Valence Aware Dictionary for Sentiment Reasoning (VADER). The time-series stationarity was determined with Augmented Dicky Fuller (ADF) Kwiatkowski Phillips Schmidt Shin (KPSS) tests and then Granger Causality testing took place. While price fluctuations seem to cause sentiment for Bitcoin, Cardano, XRP and Doge, predictability was found for Ethereum and Polkadot, based on a bullishness ratio. Finally, predictability of price returns is examined with Vector Autoregression (VAR) and highly accurate forecasts for two of the seven cryptocurrencies were achieved. More specifically, price forecasts of Ethereum’s and Polkadot’s prices reached 99.67% and 99.17% accuracy, respectively.