The goal of this paper is to ascertain with what accuracy the direction of Bitcoin price in USD can be predicted. The price data is sourced from the Bitcoin Price Index. The task is achieved with varying degrees of success through the implementation of a Bayesian optimised recurrent neural network (RNN) and a Long Short Term Memory (LSTM) network. The LSTM achieves the highest classification accuracy of 52% and a RMSE of 8%. The popular ARIMA model for time series forecasting is implemented as a comparison to the deep learning models. As expected, the non-linear deep learning methods outperform the ARIMA forecast which performs poorly. Finally, both deep learning models are benchmarked on both a GPU and a CPU with the training time on the GPU outperforming the CPU implementation by 67.7%.
Bitcoin is one of the most prominent decentralized digital cryptocurrencies. Ability to understand which factors drive the fluctuations of the Bitcoin price and to what extent they are predictable is interesting both from the theoretical and practical perspective. In this paper, we study the problem of the Bitcoin short-term volatility forecasting based on volatility history and order book data. Order book, consisting of buy and sell orders over time, reflects the intention of the market and is closely related to the evolution of volatility. We propose temporal mixture models capable of adaptively exploiting both volatility history and order book features. By leveraging rolling and incremental learning and evaluation procedures, we demonstrate the prediction performance of our model as well as studying the robustness, in comparison to a variety of statistical and machine learning baselines. Meanwhile, our temporal mixture model enables to decipher the time-varying effect of order book features on volatility. It demonstrates the prospect of our temporal mixture model as an interpretable forecasting framework over heterogeneous Bitcoin data.
In this paper, we study the ability to make the short-term prediction of the exchange price fluctuations towards the United States dollar for the Bitcoin market. We use the data of realized volatility collected from one of the largest Bitcoin digital trading offices in 2016 and 2017 as well as order information. Experiments are performed to evaluate a variety of statistical and machine learning approaches.
Cryptocurrencies return cross-predictability and technological similarity yield information on risk propagation and market segmentation. To investigate these effects, we build a time-varying network for cryptocurrencies, based on the evolution of return cross-predictability and technological similarities. We develop a dynamic covariate-assisted spectral clustering method to consistently estimate the latent community structure of cryptocurrencies network that accounts for both sets of information. We demonstrate that investors can achieve better risk diversification by investing in cryptocurrencies from different communities. A cross-sectional portfolio that implements an inter-crypto momentum trading strategy earns a 1.08% daily return. By dissecting the portfolio returns on behavioral factors, we confirm that our results are not driven by behavioral mechanisms.
M Vaidehi, Alivia Pandit, Bhaskar Jindal, Minu Kumari · 5 authors
After the boom and bust in cryptocurrencies’ prices in recent years, Bitcoin has been totally regarded as an investment asset. As it is highly volatile in nature, there has been a need for good predictions for carrying base investment decisions. Although current study has used machine learning for more accurate Bitcoin price prediction, some of them did focused on the feasibility of applying different modeling techniques to the samples that has different data structures and dimension features. To predict Bitcoin price on different frequencies after using machine learning techniques, firstly we have to classify the Bitcoin price with daily price and high-frequency price. Here, we attempt to predict Bitcoin price as accurately as possible by taking into consideration various protocols that affect the Bitcoin value. Using the provided data we would predict the sign of daily price change with highest possible accuracy. We have used Random Forest Classifier and compared with benchmark results as daily price prediction, we achieve a better performance, with the highest accuracies of the statistical methods and machine learning algorithms of 99%. my investigation in Bitcoin price prediction can be considered as a pilot study for the importance of the sample dimension in the machine learning techniques. Keywords Bitcoin, Crypto Currency, Machine Learning, Blockchain, Long Short Term Memory(LSTM), Recurrent Neural Network(RNN), Prediction
N.I. Indera, Ahmad Ihsan Mohd Yassin, Azlee Zabidi, Zairi Ismael Rizman
This paper presents a Multi-Layer Exogeneous Inputs (NARX) Bitcoin price forecasting model using the opening, closing, minimum and maximum past prices together with Moving Average (MA) technical indicators.
In this thesis, an analysis of Bitcoin, Monero price and volatility is conducted with respect to S&P500 and the VIX index. Moreover using Python, we computed correlation coefficients of nine cryptocurrencies with two different approaches: Pearson and Spearman from July 2016 -July 2018. Moreover the Pearson correlation coefficient was computed for each year from July2016 - July 2017 - July 2018. It has been concluded that in 2016 the correlation between the selected cryptocurrencies was very weak - almost none, but in 2017 the correlation increased and became moderate positive. In 2018, almost all of the cryptocurrencies were highly correlated. For example, from January until July of 2018, the Bitcoin - Monero correlation was 0.86 and Bitcoin - Ethereum was 0.82.
Bitcoin is a phenomenon that is new and there is little information on how and why it behaves as volatile as it does. This thesis uses existing data on Bitcoin’s exchange rate to estimate a model that describes the pattern and use it in a financial risk analysis. We also aim to contribute as a foundation for further studies in this field.\nThe statistical properties of the log-return of the exchange rate are analysed and it is deemed to be iid. From the eleven distributional candidates we study is the fitted skew generalised t distribution proven to represent the data best after evaluation by criteria and statistics. The estimated VaR and ES show that the rate is volatile and that the risk from investments is still high.\nThe findings show that it is necessary to describe the exchange rate with complex and flexible distributions, and even if the data shows more stability today than earlier is it important to show caution in interpretations and evaluations on the topic.\nKeywords: Bitcoin, cryptocurrencies, statistical distributions, statistical analysis, exchange rate, modelling
The aim of the paper is to fit a regression model which can be used commonly for the four important crypto currencies: Bitcoin, Litecoin, Ethereum and Ripple to predict the prices. The data has information over the past six years regarding price, transaction volume, transaction count, exchange volume, generated coins etc of these currencies. Understanding the dynamics of crypto currency market can help to a certain extent to take wise investment decisions. Among the variables under consideration the study revealed that transaction volume can be used as an influencing variable to fit a quadratic regression model and predict the prices of the crypto currencies.
This research examines existence of Bitcoin bubble in the year of 2017 by studying time series data within four main paradigms of the modern bubble theory. All four models suggest that Bitcoin closing price was overstated during the period. The analysis also detects Bitcoin features that could lead to the behavioral bias on the cryptocurrency market. The study suggests that investors’ behavior on the crypto market should be investigated more within behavioral finance.
This paper attempts to establish that some inherent features of the Bitcoin price can be exploited to produce better forecast results for stock prices. It does so by constructing predictive models for stock prices of G7 countries with symmetric and asymmetric prices of Bitcoin. The underlying statistical properties of Bitcoin prices such as persistence and conditional heteroscedasticity are captured in the estimation process using the Westerlund and Narayan (2015) estimator that allows for such effects in forecasting. There are two striking findings from the analysis. First, the results suggest that accounting for asymmetries is more likely to enhance the predictive power of Bitcoin in forecasting stock prices regardless of the data sample and forecast horizon. Secondly, the Bitcoin-based predictive model for stock prices, particularly the asymmetric variant, outperforms the Fractionally Integrated Autoregressive Moving Average (ARFIMA) model. While there are concerns as to whether the cryptocurrencies are veritable substitutes to the conventional financial assets, their close link with the developed stock exchanges such as those in the G7 countries suggests that they share some common characteristics such as news effects [asymmetries] which can be exploited when forecasting the behaviour of stock prices.
The cryptocurrency market has become increasingly accessible and significant to the financial markets. This is understood by not only major financial firms, governments, and investors, but also the individual market participants globally. We delve into the history of cryptocurrency to begin our examination of the Bitcoin, Ethereum and Litecoin. Understanding the circumstances of their humble beginning, the purpose it served, and the path of their evolution, helps us to create a fuller understanding of its functions, its limitations, and the drivers of its value. This enables us to identify key market factors and variables for deployment within a robust approach for pricing and product offerings associated with Bitcoin, Ethereum and Litecoin. In order to fully capture the volume, variety, and velocity of data associated with these cryptocurrencies, the use of machine learning can provide an advantageous approach to model development for cryptocurrency pricing. This paper provides the development of a promising initial prototype pricing model for Bitcoin, Ethereum and Litecoin. Our proposed pricing models resulted in an average 7% difference between actual and predicted price for Bitcoin and Ethereum, and a 4% difference for Litecoin along a timeline, through the use of machine learning and deep learning, artificial neural networks using the contributing factors of key variables and how they influence and capture pricing and investor behaviour. We also identify theinclusion of additional datasets, such as sentiment market data into the model, along with larger exploration of Blockchain and raw transaction mining to increase the accuracy and forecasting ability of the model.
Usman Amjad, Tahseen Ahmed, Humera Tariq, Amir Hussain
Quantum computing has emerged as a new dimension with various applications in different fields like robotic, cryptography, uncertainty modeling etc. On the other hand, nature inspired techniques are playing vital role in solving complex problems through evolutionary approach. While evolutionary approaches are good to solve stochastic problems in unbounded search space, predicting uncertain and ambiguous problems in real life is of immense importance. With improved forecasting accuracy many unforeseen events can be managed well. In this paper a novel algorithm for Fuzzy Time Series (FTS) prediction by using Quantum concepts is proposed in this paper. Quantum Evolutionary Algorithm (QEA) is used along with fuzzy logic for prediction of time series data. QEA is applied on interval lengths for finding out optimized lengths of intervals producing best forecasting accuracy. The algorithm is applied for forecasting Taiwan Futures Exchange (TIAFEX) index as well as for Bitcoin crypto currency time series data as a new approach. Model results were compared with many preceding algorithms.
Cryptocurrencies are digital currencies that have garnered significant investor attention in the financial markets.The aim of this project is to predict the daily price, particularly the daily high and closing price, of the cryptocurrency Bitcoin.This plays a vital role in making trading decisions.There exist various factors which affect the price of Bitcoin, thereby making price prediction a complex and technically challenging task.To perform prediction, we trained temporal neural networks such as time-delay neural networks (TDNN) and recurrent neural networks (RNN) on historical time seriesthat is, past prices of Bitcoin over several years.Features such as the opening price, highest price, lowest price, closing price, and volume of a currency over several preceding quarters were taken into consideration so as to predict the highest and closing price of the next day.We designed and implemented TDNNs and RNNs using the NeuroSolutions artificial neural network (ANN) development environment to build predictive models and evaluated them by computing various measures such as the MSE (mean square error), NMSE (normalized mean square error), and r (Pearson's correlation coefficient) on a continuation of the training data from each time series, held out for validation.
In response to Molnár and Thies (2018) demonstrating that the price data of Bitcoin contained structural breaks, we identify the optimal number of states for a Markov regime-switching (MRS) model to capture the regime heteroskedasticity of Bitcoin. We determined that the restricted 5-state MRS model provided the best goodness-of-fit scores (-AIC, -BIC, -HQIC) for the fitted sample. In addition, we found evidence of stylised characteristics in the price data of Bitcoin, namely: volatility clustering; volatility jumps; asymmetric volatility transitions; and the persistence of shocks.
This paper studies Granger Causality relations between Bitcoin and 5 stock market indexes which are Japan, Russia, South Korea, Sweden and the United States. The time-period examined is from 2013 to 2017 and all the tests are conducted based on daily data. We analyze this in three different periods, last 5 years (2013-2017), in 2017 and last 3 months of 2017. To estimate the relationship, we use unit root test and Augmented Dickey-Fuller, Lagrange Multiplier, Johansen Cointegration Test and finally Granger Causality Test. After the tests, countries have a same integrated order that exhibits a long-run relationship. In causality, except for Russia, each country has affected the Bitcoin prices and being affected in a different period, especially in the last 3 months of 2017, the impact and popularity of Bitcoin affect too much the stock market in the short-run. As a result, the causation between Bitcoin and stock market indexes shows impact statistically significant in the 2017 year. The importance of cryptocurrency and popularity not as much as hype like late 2017 in 2018, but we think that cryptocurrencies are one of the major currencies that affect economical world very deeply.
The purpose of this study is to develop robust estimation of association between two types of crypto-currencies namely Bitcoin and Ethereum. Daily data of crypto-currencies are collected from https://coinmarketcap.com. The period for data analysis is started from January 2017 until October 2018. The value of mean return for Bitcoin is 13.18 %. Meanwhile, the value of mean return for Ethereum is 27.85 %. The standard deviation for Bitcoin is 30.27 % and Ethereum is 64.24 %. Then, this study performed Person product moment coefficient analysis to evaluate the correlation between these two crypto-currencies. Result indicates the association coefficient value is 0.50. The correlation shows there is strong positive correlation between Bitcoin return and Ethereum return. As conclusion, there is significant relationship between Bitcoin and Ethereum return data with strong positive correlation (r = 0.503, n = 21, p =0.020).The significant of this study is to help investors to make better decision in selecting appropriate investment portfolio for their investment fund that contributes better return and lower risk.
Abstract This study investigates the profitability of an algorithmic trading strategy based on training SVM model to identify cryptocurrencies with high or low predicted returns. A tail set is defined to be a group of coins whose volatility-adjusted returns are in the highest or the lowest quintile. Each cryptocurrency is represented by a set of six technical features. SVM is trained on historical tail sets and tested on the current data. The classifier is chosen to be a nonlinear support vector machine. The portfolio is formed by ranking coins using the SVM output. The highest ranked coins are used for long positions to be included in the portfolio for one reallocation period. The following metrics were used to estimate the portfolio profitability: %ARC (the annualized rate of change), %ASD (the annualized standard deviation of daily returns), MDD (the maximum drawdown coefficient), IR1, IR2 (the information ratio coefficients). The performance of the SVM portfolio is compared to the performance of the four benchmark strategies based on the values of the information ratio coefficient IR1, which quantifies the risk-weighted gain. The question of how sensitive the portfolio performance is to the parameters set in the SVM model is also addressed in this study.
Predicting currency prices remains a difficult endeavour. Investors are continually seeking new ways to extract \nmeaningful information about the future direction of price changes. Recently, cryptocurrencies have attracted \nhuge attention due to their unique way of transferring value as well as its value as a hedge. A method proposed \nin this project involves using data mining techniques: mining text documents such as news articles and tweets \ntry to infer the relationship between information contained in such items and cryptocurrency price direction. \nThe Long Short-Term Memory Recurrent Neural Network (LSTM RNN) assists in creating a hybrid model \nwhich comprises of sentiment analysis techniques, as well as a predictive machine learning model. The success \nof the model was evaluated within the context of predicting the direction of Bitcoin price changes. Findings \nreported here reveal that our system yields more accurate and real-time predictions of Bitcoin price fluctuations \nwhen compared to other existing models in the market.
Zhenghui Li, Hao Dong, Zhehao Huang, Pierre Failler
The rapid development of VFAs allows investors to diversify their choices of investment products. In this paper, we measure the return risk of VFAs based on GARCH-type model. By establishing a Markov regime-switching Regression (MSR) Model, we explore the asymmetric effects of speculation, investor attention, and market interoperability on return risks in different risk regimes of VFAs. The results show that the influences of speculation and investor attention on the risks of VFAs are significantly positive at all regimes, while market interoperability only admits a positive impact on risk under high risk regime. All of the three factors exert asymmetric effects on risks in different regimes. Further study presents that the risk regime-switching also shows asymmetric characteristic but the medium risk regime is more stable than any others. Therefore, transactions of investors and arbitrageurs are monitored by certain policies, such as limiting the number of transactions or restricting the trading amount at high risk regime. However, when return risk is low, it will return to a medium level if we encourage investors to access.
We study investor sentiment on a non-classical asset such as cryptocurrency using machine learning methods. We account for context-specific information and word similarity by using efficient language modelling tools such as construction of featurized word representations (embeddings) and recursive neural networks (RNNs). We apply these tools for sentence-level sentiment classification and sentiment index construction. This analysis is performed on a novel dataset of 1220K messages related to 425 cryptocurrencies posted on a microblogging platform StockTwits during the period between March 2013 and May 2018. Both in- and out-of-sample predictive regressions are run to test significance of the constructed sentiment index variables. We find that the constructed sentiment indices are informative regarding returns' and volatility predictability of the cryptocurrency market index.