Deep learning methods have achieved significant success in various applications, including trend signal prediction in financial markets. However, most existing approaches only utilize price action data. In this paper, we propose a novel system that incorporates multiple data sources and market correlations to predict the trend signal of Ethereum cryptocurrency. We conduct experiments to investigate the relationship between price action, candlestick patterns, and Ethereum-Bitcoin correlation, aiming to achieve highly accurate trend signal predictions. We evaluate and compare two different training strategies for Convolutional Neural Networks (CNNs), one based on transfer learning and the other on training from scratch. Our proposed 1-Dimensional CNN (1DCNN) model can also identify inflection points in price trends during specific periods through the analysis of statistical indicators. We demonstrate that our model produces more reliable predictions when utilizing multiple data representations. Our experiments show that by combining different types of data, it is possible to accurately identify both inflection points and trend signals with an accuracy of 98%.
This research assesses the prediction of Bitcoin prices using the autoregressive integrated moving average (ARIMA) and long-short-term memory (LSTM) models. We forecast the price of Bitcoin for the following day using the static forecast method, with and without re-estimating the forecast model at each step. We take two different training and test samples into consideration for the cross-validation of forecast findings. In the first training sample, ARIMA outperforms LSTM, but in the second training sample, LSTM exceeds ARIMA. Additionally, in the two test-sample forecast periods, LSTM with model re-estimation at each step surpasses ARIMA. Comparing LSTM to ARIMA, the forecasts were much closer to the actual historical prices. As opposed to ARIMA, which could only track the trend of Bitcoin prices, the LSTM model was able to predict both the direction and the value during the specified time period. This research exhibits LSTM's persistent capacity for fluctuating Bitcoin price prediction despite the sophistication of ARIMA.
Grzegorz Dudek, Piotr Fiszeder, Paweł Kobus, Witold Orzeszko
Forecasting cryptocurrency volatility can help investors make better-informed investment decisions in order to minimize risks and maximize potential profits. Accurate forecasting of cryptocurrency price fluctuations is crucial for effective portfolio management and contributes to the stability of the financial system by identifying potential threats and developing risk management strategies. The objective of this paper is to provide a comprehensive study of statistical and machine learning methods for predicting daily and weekly volatility of the following four cryptocurrencies: Bitcoin, Ethereum, Litecoin, and Monero. Several models and forecasting methods are compared in terms of their forecasting accuracy, i.e., HAR (heterogeneous autoregressive), ARFIMA (autoregressive fractionally integrated moving average), GARCH (generalized autoregressive conditional heteroscedasticity), LASSO (least absolute shrinkage and selection operator), RR (ridge regression), SVR (support vector regression), MLP (multilayer perceptron), FNM (fuzzy neighbourhood model), RF (random forest), and LSTM (long short-term memory). The realized variance calculated from intraday returns is used as the input variable for the models. In order to assess the predictive power of the models considered, the model confidence set (MCS) procedure is applied. Our experimental results demonstrate that there is no single best method for forecasting volatility of each cryptocurrency, and different models may perform better depending on the specific cryptocurrency, choice of the error metric and forecast horizon. For daily forecasts, the method that is always found in a set of best models is linear SVR, while for weekly forecasts, there are two such methods, namely FNM and RR. Furthermore, we show that simple linear models such as HAR and ridge regression, perform not worse than more complex models like LSTM and RF. The research provides a useful reference point for the development of more sophisticated models.
Anand Shankar Raja M., Benita Priyadarshini D., Janani Govindaraj, Saket Agarwal
Cryptocurrency is a commonly used term in the current world, and the COVID-19 pandemic has indirectly increased the awareness and the investor base for cryptocurrencies. Various research has been conducted to understand the complex working structure of these investment options and to analyse the volatile nature of cryptocurrencies. There are multiple factors and triggers that impact the price movements in the crypto market. Classifying these factors would help streamline the process of analysing these factors for further studies. These factors cause both positive and negative impacts on the price fluctuations. Classifying the major factors under the period of impact will help understand each factor's role in the market. This classification would help in the diagnostic and prescriptive analysis of cryptocurrencies. In this research, well-cited and published research papers, journals, and articles have been studied to classify some of the major factors affecting cryptocurrencies carefully. A model has been created to easily comprehend the classification of factors based on time of impact. This model simplifies the understanding of the factors and would help conduct further analysis on these factors.
In recent decades, data analytics has become increasingly involved in people's daily lives. Machine learning, an important part of data analysis, has also been used in the financial sector. Contemporarily, the high volatility feature of cryptocurrencies has attracted lots of investors, which also brings lots of difficulty to predict and analyze. In fact, the price of cryptocurrencies can also be forecasted based on machine learning. This paper uses historical data of Bitcoin, Ethereum and Dogecoin as inputs to predict the future value based on the LSTM. LSTM model can learn the long-term dependencies in data. According to the analysis, mean absolute error calculate the average size of the error in a set of predictions, regardless of its direction. The results produced can roughly predict the future trends of these three cryptocurrencies. This paper combines the fields of machine learning and finance to predict the future value of cryptocurrencies. These results shed light on guiding further exploration of predicting cryptocurrency assets valuation based on LSTM model.
The study examines the problem of modeling and forecasting the price dynamics of crypto currencies. We use machine learning techniques to forecast the price of crypto currencies. The FB Prophet time series model and the LSTM recurrent neural network were selected to implement the study. Using the example of data from Binance (the most popular exchange in Ukraine) for the period from 06.07.2020 to 01.04.2023, prices for Bitcoin, Ethereum, Ripple, and Dogecoin were modeled and forecasted. The recurrent neural network of long-term memory showed significantly better results in forecasting according to the RMSE, MAE, and MAPE criteria, compared to the Naïve model, the traditional ARIMA model, and the FB Prophet results.
Since it was first founded by Satoshi Nakamoto in 2008, cryptocurrencies have attracted the attention of investors significantly. Until now, many investors have invested their money in cryptocurrencies. This study aims to prove and analyze the effect of volatility, liquidity and world oil price indices on cryptocurrency returns. It also examines whether the occurrence of market efficiency is weak or not. This study aims to prove and analyze weak efficiency market and the impact of liquidity, volatility and the world oil price index on the return of the cryptocurrency market. This study uses quantitative methods and secondary data, so that the number of samples taken are 3 types of cryptocurrencies (Bitcoin, Ethereum and Binance Coin) which are listed on coinmarketcap.com and investing.com, as well as the oil price index taken from investing.com The analytical methods used in this study are series correlation test and runs test as well as multiple linear regression analysis. The results of this study indicate that volatility and liquidity affect the efficiency of the cryptocurrency market while the oil price index has no effect. The results of autocorrelation test on return itself prove that these coins have weak form efficiency. Keywords: Market Efficiency, Cryptocurrency, Liquidity, Oil Price Index, Volatility
Abstract: Cryptocurrency, also known as crypto, is any digital or virtual currency that uses cryptography to safeguard transactions and circulates without the authority of a central bank. Bitcoin, the first and most widely used decentralized cryptocurrency, was introduced in 2009. After a few years of unrivalled dominance, it lost its monopoly in 2011, when the first competitive alternative currencies arose. As of November 2022, there are almost 21,000 cryptocurrencies in circulation. Because there is no government credit backup, cryptocurrency prices are typically volatile. The cost of one Bitcoin rose from zero at its debut in 2009 to $13 in 2013 and then to $68789 in 2021, with numerous shifts and fluctuations along the way. The accurate forecasting of the Bitcoin price is critical for investors to make decisions and for governments to create regulatory laws. This paper examines the ability of the models - Prophet, Long Short-Term Memory (LSTM), and eXtreme Gradient Boosting (XGBoost) to predict the price of Bitcoin reliably. Using the performance metrics like RMSE, each model was thoroughly trained and tested to discover which one operates more efficiently. After examining the price of Bitcoin from 2012 to 2021, we concluded that the Long Short-Term Memory (LSTM) model proves to be the most efficient when dealing with variable and difficult-topredict data such as Bitcoin values since it portrays promising results in comparison
Muhammad Zakhwan, Mohamed Rafik, Noraisyah Mohamed Shah, Anis Salwa Binti Mohd Khairuddin
Cryptocurrency is branded as a digital currency, an alternative exchange currency system with significant ramifications for the economies of rising nations and the global economy. In recent years, cryptocurrency has infiltrated almost all financial operations; hence, cryptocurrency trading is frequently recognised as one of the most popular and promising means of profitable investment. Lately, with the exponential growth of cryptocurrency in-vestments, many Alternative Coins (Altcoins) resurfaced as to mimic the fiat currency. Altcoins prediction, as the name suggests the alternative coins from the traditional cryptocurrency which is Bitcoin (BTC). There are several methods to forecast cryptocurrency prices namely Technical Analysis and Fundamental Analysis which has been widely used in forecasting fiat and stock prices. With the emergence of Artificial Intelligence (AI), Machine Learning and Deep Learning algorithms provide a different perspective on how investors can estimate the trend or the movement of prices. In this thesis, as cryptocurrency price are time-dependent, Recur-rent Neural Network (RNN) is presented due to RNN’s nature that is well suited for Time Series Analysis (TSA). The topology of proposed RNN model consists of 3 stages which are model groundwork, model development and testing and optimisation. The RNN architecture are extended to two different models specifically Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU). There are 4 hyperparameters that will affect the accuracy of the deep learning model in predicting cryptocurrency price. Hyperparameters tuning set the basis of optimising the model to improve the accuracy of cryptocurrency prediction. Hyperparameters listed in this project are limited to number of epochs, adaptive optimisation algorithm, dropout rate, and batch size. Next, the models are tested with data of different coins listed in the cryptocurrency market with different input features to find out the effect on the accuracy and robustness of the model in predicting the cryptocurrency price. This research demonstrates that GRU has the best accuracy in forecasting the cryptocurrency prices based on the values of Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and Executional Time, scoring 2.2201, 0.8076 and 200s using intra-day trading strategy Open, High, Low, Close Price (OHLC) as input features.
Abstract: Predicting future events is difficult, particularly with regards to cryptocurrency, where the media, influential people and governments have a sharp and vital impact on worth. Cryptocurrency market analysis is a method through which the realworld data of the cryptocurrency market is used to predict where it will go next. If foretold accurately, it helps investors to invest when the value is low (purchasing in bulk when the price is dipping) and sell once it's high so as to gain a profit. This research provides two machine learning algorithms which are Long Short-Term Memory (LSTM) and Linear Regression for predicting the values of six different types of crypto currencies such as Bitcoin (BTC), Dash coin (DASH), Lite coin (LTC), Dogecoin (DOGE), Ethereum (ETH), and Monero (XMR). The accuracy of the models is analyzed using mean squared error.
Formerly, predictions of time series were usually based on numbers. In some cases, however, one has to make predictions based on linguistic information. In this paper, we experiment with time series forecasting based on textual information. We combine sentiment analysis with time series forecasting to predict fluctuations in variables that change over time. Due to the time-sensitivity of the target variable, models from the Time Series Prediction Database (tspDB) are used to infer the value of Bitcoin. We implemented an innovative architecture that accepts text as input and produces numerical predictions for certain values. The market value of bitcoin was used to verify the applicability of the architecture. The program inputs Twitter posts and outputs the market value of bitcoin for the next few days. The program first calculates a sentiment score (which reflects social media confidence in the price of bitcoin) and then makes a multivariate time series prediction of the price of bitcoin. The results can provide some insight into how the price of bitcoin fluctuates. The purpose of using Bitcoin as our primary benchmark is that we want to demonstrate a sound application of the combination of NLP and tspDB, and while Bitcoin price is not an objective variable, the method we apply can also be used in other situations that contain linguistic information as input.
With the rapid growth of technology, cryptocurrency like Bitcoin is attracting more and more attention. Its high volatility in prices creates many difficulties for predicting and there has been much work on this. This paper aims to provide a comparison of various machine learning models like linear regression, SVM, random forest, and neural networks for predicting the directions for Bitcoin close prices. The dataset used is from Jan 2012 to March 2021 and all four prices are used for predictions: Close, Open, High, and Low. Two different methods are used to fit the different types of machine learning algorithms: for regressors, close price predictions are first done and then construct in the predicted direction; for classifiers, direction predictions are done directly. Accuracy is used to do the comparison, which is the percentage of correct direction predictions made via the algorithm. It is shown that LSTM, a neural network algorithm generates the highest accuracy of about 58% and the random forest classifier has the lowest accuracy of about 55.47%.
Recent years have witnessed the rapid development of bitcoin as the first digital currency. Considering the advantages of bitcoin for both individuals and society, the price prediction of bitcoin is a hot topic. However, there remain two main challenges to be addressed in this task. Firstly, because bitcoin is vulnerable to the attitudes of the investors, incorporating the sentiment semantics from the social media into the prediction is challenging. Secondly, it is intractable to predict extreme volatility of the bitcoin price. To tackle the above challenges, this paper proposes to incorporate sentiment and temporal information simultaneously. For the first challenge, this paper employs external unsupervised corpus to conduct the domain-specific post-pretraining on the off-the-shelf language model. And the sentiment analysis on the tweets is done to obtain the scores. For the second one, Long Short Term Memory (LSTM) network is leveraged to joint model the temporal price data and the sentiment scores, thus deriving the final predictions. Experiments on the real data show that compared with the single-layer LSTM model, the model in this paper works better, which provides help for investors to specify trading strategies and also provides implications for government agencies that are developing digital currencies.
Bitcoin is the most famous digital currency in the world and has become an investment asset. Prediction is one of the important matters in the investment market. In the economic field, there are different studies on the reasons for the price change of Bitcoin and how to predict the price trend of Bitcoin or how Bitcoin studies the market. Therefore, for Bitcoin, predicting the trend of Bitcoin price can effectively help Bitcoin investors. Data from www. Coingecko, the price of bitcoin is sorted according to the time sequence. Using the time series model, the change of bitcoin price in a specific period which is from 28 April 2013 to 22 August 2022 is calculated to predict the future trend of bitcoin price. Data preprocessing includes attributes removal, stationary test, and differencing. In predicting the price of Bitcoin, the ARIMA method that can produce high accuracy in short-term prediction is adopted. Use prediction test AIC and Check the residuals to select the best prediction model among the candidate models. The results of model testing show that AIC of ARIMA (5,1,2) is the smallest among all candidate models, and the results of residual check also show that ARIMA (5,1,2) model is the best model for predicting four periods.
In this paper, Autoregressive Integrated Moving Average (ARIMA) model is used for analysing and forecasting the adjusted closing price of bitcoin. The whole dataset used is daily bitcoin closing price dating from Jan 2017 to Sep 2022. However, for testing the performance of the ARIMA model, the dataset is divided into two parts: the ARIMA model is built on training set and later using the test set to check the accuracy of the prediction. Two models, namely ARIMA (5, 2, 1) and ARIMA (0, 2, 2) are selected and comparations between them are made. ARIMA (5, 2, 1) is chosen with stepwise selection and approximated information criteria while ARIMA (0, 2, 2) is without stepwise selection and the information criteria is not approximated. Both pass the residual test and ARIMA (0, 2, 2) is slightly better according to AIC, AICc and BIC. Later, the predicting accuracy of the two models for different forecasting periods (5-day, 10-day, and long-term forecasting) are compared. It is not surprising that ARIMA performs better while making short term prediction. The 5-day and 10-day forecast works well while the long-term forecast is of limited practical value.
The time series movements of Bitcoin prices are commonly characterized as highly nonlinear and volatile in nature across economic periods, when compared to the characteristics of traditional asset classes, such as equities and commodities. From a risk management perspective, such behaviors pose challenges, given the difficulty in quantifying and modeling Bitcoin’s price volatility. In this study, we propose hybrid analytical techniques that combine the strengths of the non-stationary properties of Generalized Autoregressive Conditional Heteroskedasticity (GARCH) models with the nonlinear modeling capabilities of deep learning algorithms, such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM) algorithms with single, double, and triple layer network architectures to forecast Bitcoin’s realized price volatility. Our findings, both in-sample and out-of-sample, show that such hybrid models can generate accurate forecasts of Bitcoin’s price volatility.
With the rise of Blockchain technology, the cryptocurrency market has been gaining significant interest. In particular, the number of cryptocurrency traders and the market capitalization have grown tremendously. However, predicting cryptocurrency price is very challenging and difficult due to the high price volatility. In this paper, we propose a classification machine learning approach in order to predict the direction of the market (i.e., if the market is going up or down). We identify key features such as Relative Strength Index (RSI) and Moving Average Convergence Divergence (MACD) to feed the machine learning model. We illustrate our approach through the analysis of Bitcoin close price. We evaluate the proposed approach via different simulations. Particularly, we provide a backtesting strategy. The evaluation results show that the proposed machine learning approach provides buy and sell signals with more than 86% accuracy.
Currently, the most popular cryptocurrency is bitcoin. Predicting the future value of bitcoin can help investors to make more educated decisions and to provide authorities with a point of reference for evaluating cryptocurrency. The novelty of the proposed prediction models lies in the use of artificial intelligence to identify movement cryptocurrency prices, particularly bitcoin prices. A forecasting model that can accurately and reliably predict the market’s volatility and price variations is necessary for portfolio management and optimization in this continually expanding financial market. In this paper, we investigate a time series analysis that makes use of deep learning to investigate volatility and provide an explanation for this behavior. Our findings have managerial ramifications, such as the potential for developing a product for investors. This can help to expand upon our model by adjusting various hyperparameters to produce a more accurate model for predicting the price of cryptocurrencies. Another possible managerial implication of our findings is the potential for developing a product for investors, as it can predict the price of cryptocurrencies more accurately. The proposed models were evaluated by collecting historical bitcoin prices from 1 January 2021 to 16 June 2022. The results analysis of the GRU and MLP models revealed that the MLP model achieved highly efficient regression, at R = 99.15% during the training phase and R = 98.90% during the testing phase. These findings have the potential to significantly influence the appropriateness of asset pricing, considering the uncertainties caused by digital currencies. In addition, these findings provide instruments that contribute to establishing stability in cryptocurrency markets. By assisting asset assessments of cryptocurrencies, such as bitcoin, our models deliver high and steady success outcomes over a future prediction horizon. In general, the models described in this article offer approximately accurate estimations of the real value of the bitcoin market. Because the models enable users to assess the timing of bitcoin sales and purchases more accurately, they have the potential to influence the economy significantly when put to use by investors and traders.
Wajid Shakeel Ahmed, Ahsan Mehmood, Talha Sheikh, Allah Bachaya
This paper investigated the relationship between cryptocurrencies and emerging stock market indices using fractional integration and co-integration technique. Particularly, fractional integration is applied to examine stochastic properties of individual assets and fractional cointegration to analyse bivariate connectedness. Our findings unveil the absence of mean reversion in majority cases which indicates high persistence in series. Furthermore, bivariate analysis reveals disconnection between cryptocurrencies prices and stock indices. Surprisingly, a different picture emerges on using conditional volatility instead of prices. Like, conditional volatility-based estimation uncovers evidence of mean reversion in univariate analysis as expected. There is some evidence of cointegration on volatility grounds between cryptocurrencies and emerging stock market indices. Our findings implies that investment decision regarding digital currencies should be taken cautiously. As cryptocurrencies are extremely volatile with high degree of persistence which can make them counterproductive.
Everyone is eager for high yield and low risk. In this research, we use Markowitz's investment theory and Monte Carlo simulation to find the optimal investment portfolio and then study the impact of adding Bitcoin to the traditional investment portfolio on the cumulative rate of return. Our results show that the return performance of the investment portfolio with Bitcoin is better than that of the traditional investment portfolio. Moreover, despite the impact of COVID-19 on the global economy and the Federal Reserve's quantitative easing policy, it is beneficial for investors to include Bitcoin in their portfolio allocation.
This study investigates mean and volatility spillover effects among eight major cryptocurrencies; Bitcoin, Ethereum, Litecoin, Ripple, Stellar, Bitcoin Cash, Cardano and EOS utilizing VAR-BEKK-GARCH model. The results point out that there are bidirectional and unidirectional spillover effects among these major cryptocurrencies. Moreover, the findings indicate that some cryptocurrencies are the transmitter, while others act as a receiver and among all, Litecoin is the highest transmitter, and Stellar is the only one that acts as a receiver. The interdependence among cryptocurrencies supports that they are becoming more integrated and thereby, provides important investment strategies for investors and policy implications for regulators.