R. Sujatha, V Mareeswari, Jyotir Moy Chatterjee, Abd Allah A. Mousa · 5 authors
Bitcoin is a decentralized digital currency without a central bank or single administrator sent from user to user on the peer-to-peer bitcoin blockchain network without intermediaries' need. In this Bitcoin trend analysis work, initial attributes are considered from five sectors based on financial, social, token, network, and that count to thirteen attributes. The thirteen attributes considered are price, volume, market cap, a mean dollar invested age, social volume, social dominance, development activity, transaction volume, token age consumed, token velocity, token circulation, market value to realized value, and realized cap. We apply the attribute selection and trend analysis mapped with potential seven attributes: Price, Volume, Market Cap, Social Dominance, Development Activity, Market Value to Realized Value & Realized Cap. We have conducted Nonlinear Autoregressive with External Input analysis considering seven attributes. The work employed three training algorithms to train a neural network as Levenberg-Marquard, Bayesian Regularization, and Scaled Conjugate Gradient algorithm. The Error histogram and regression plots results indicate that the Bayesian Regularized Neural Network is showing good performance and thus provides a better forecast.
During recent developments, cryptocurrency has become a famous key factor in financial and business opportunities. However, the cryptocurrency investment is not visible regarding the market’s inconsistent aspect and volatility of high prices. Due to the real-time prediction of prices, the previous approaches in price prediction doesn’t contain enough information and solution for forecasting the price changes. Based on the mentioned problems in cryptocurrency price prediction, we proposed a machine learning-based approach to price prediction for a financial institution. The proposed system contains the blockchain framework for secure transaction environment and Reinforcement Learning algorithm for analysis and prediction of price. The main focus of this system is on Litecoin and Monero cryptocurrencies. The results show the presented system accurate the performance of price prediction higher than another state-of-art algorithm.
The paper examines the influence of investor sentiment based on news headlines on the Cryptocurrency Market Index and ten individual cryptocurrency returns. We capture investors’ sentiment from cryptocurrency-specific news headlines. We use a lexicon-based Natural Language Processing (NLP) technique to construct a unique sentiment indicator, and the sentiment scores are generated using two financial dictionaries: Henry(2008)(HE) and Loughran and Mcdonald(2011)(LM). The findings of the study show that news sentiment has a significant impact on cryptocurrency returns. When the investors’ sentiment is optimistic or bullish, the cryptocurrency market experiences herding behaviour, leading to an increase in prices. The diverse and heterogeneous nature of the various cryptocurrencies causes each individual cryptocurrency to respond differently to sentiment. Further, we see that sentiment has a more pronounced impact on young, small, and volatile cryptocurrencies. Our study is among the few studies that use cryptocurrency-specific news headlines rather than news bodies to build a news sentiment indicator. JEL codes: E49, G14, G15
This paper discusses the problems of short-term forecasting of cryptocurrency time series using a supervised machine learning (ML) approach. For this goal, we applied two of the most powerful ensemble methods including Random Forests (RF) and Stochastic Gradient Boosting Machine (SGBM). As the dataset was collected from daily close prices of three of the most capitalized coins: Bitcoin (BTC), Ethereum (ETH) and Ripple (XRP), and as features we used past price information and technical indicators (moving average). To check the effectiveness of these models we made an out-of-sample forecast for selected time series by using the one step ahead technique. The accuracy rate of the forecasted prices by using RF and GBM were calculated. The results verify the applicability of the ML ensembles approach for the forecasting of cryptocurrency prices. The out of sample accuracy of short-term prediction daily close prices obtained by the SGBM and RF in terms of Mean Absolut Percentage Error (MAPE) for the three most capitalized cryptocurrencies (BTC, ETH, and XRP) were within 0.92-2.61 %.
Blockchain technology is becoming increasingly popular because of its applications in various fields. It gives an edge over the traditional centralized methods as it provides decentralization, immutability, integrity, and anonymity. The most popular application of this technology is cryptocurrencies, which showed a massive rise in their popularity and market capitalization in recent years. Individual investors, big institutions, and corporate firms are investing heavily in it. However, the crypto market is less stable than traditional commodity markets. It can be affected by many technical, sentimental, and legal factors, so it is highly volatile, uncertain, and unpredictable. Plenty of research has been done on various cryptocurrencies to forecast accurate prices, but the majority of these approaches can not be applied in real-time. Motivated from the aforementioned discussion, in this paper, we propose a deep-learning-based hybrid model (includes Gated Recurrent Units (GRU) and Long Short Term Memory (LSTM)) to predict the price of Litecoin and Zcash with inter-dependency of the parent coin. The proposed model can be used in real-time scenarios and it is well trained and evaluated using standard data sets. Results illustrate that the proposed model forecasts the prices with high accuracy compared to existing models.
Recent studies in big data analytics and natural language processing develop automatic techniques in analyzing sentiment in the social media information. In addition, the growing user base of social media and the high volume of posts also provide valuable sentiment information to predict the price fluctuation of the cryptocurrency. This research is directed to predicting the volatile price movement of cryptocurrency by analyzing the sentiment in social media and finding the correlation between them. While previous work has been developed to analyze sentiment in English social media posts, we propose a method to identify the sentiment of the Chinese social media posts from the most popular Chinese social media platform Sina-Weibo. We develop the pipeline to capture Weibo posts, describe the creation of the crypto-specific sentiment dictionary, and propose a long short-term memory (LSTM) based recurrent neural network along with the historical cryptocurrency price movement to predict the price trend for future time frames. The conducted experiments demonstrate the proposed approach outperforms the state of the art auto regressive based model by 18.5% in precision and 15.4% in recall.
Ahmed M. Khedr, Ifra Arif, Pravija Raj P V, Magdi El‐Bannany · 6 authors
Abstract Cryptocurrencies are decentralized electronic counterparts of government‐issued money. The first and best‐known cryptocurrency example is bitcoin. Cryptocurrencies are used to make transactions anonymously and securely over the internet. The decentralization behavior of a cryptocurrency has radically reduced central control over them, thereby influencing international trade and relations. Wide fluctuations in cryptocurrency prices motivate the urgent requirement for an accurate model to predict its price. Cryptocurrency price prediction is one of the trending areas among researchers. Research work in this field uses traditional statistical and machine‐learning techniques, such as Bayesian regression, logistic regression, linear regression, support vector machine, artificial neural network, deep learning, and reinforcement learning. No seasonal effects exist in cryptocurrency, making it hard to predict using a statistical approach. Traditional statistical methods, although simple to implement and interpret, require a lot of statistical assumptions that could be unrealistic, leaving machine learning as the best technology in this field, being capable of predicting price based on experience. This article provides a comprehensive summary of the previous studies in the field of cryptocurrency price prediction from 2010 to 2020. The discussion presented in this article will help researchers to fill the gap in existing studies and gain more future insight.
Nishant Jagannath, Tudor Barbulescu, Karam M. Sallam, Ibrahim Elgendi · 8 authors
The Ethereum blockchain generates a significant amount of data due to its intrinsic transparency and decentralized nature. It is also referred to as on-chain data and is openly accessible to the world. Moreover, the on-chain data is timestamped, integrated, and validated into an open ledger. This important blockchain feature enables us to assess the network’s health and usage. It serves as a massive data warehouse for complex prediction algorithms that can effectively detect systemic trends and forecast future behavior. We adopt a quantitative approach using a subset of these metrics to determine the network’s true monetary value by developing a Long Short-Term Memory Recurrent Neural Network (LSTM-RNN) with the metrics most closely associated with the price as inputs. Since several hyperparameters regulate the learning process in an RNN, they are highly sensitive to their values. It is thus critical, to select optimal hyperparameters so that the training is quick and effective. Determining the optimal parameters of an RNN model is a tedious and complex process. Hence, previous studies have developed several self-adaptive approaches to determine the optimal values for various parameters effectively. However, none of the prior studies explore self-adaptive algorithms in deep learning models in conjunction with on-chain data to predict cryptocurrency prices. In this paper, we propose three self-adaptive techniques, each of which converges on a set of optimal parameters to predict the price of Ethereum accurately. We compare our results to a traditional LSTM model. Our approach exhibits 86.94% accuracy while maintaining a minimum error rate.
Most studies in the Bitcoin literature are focused on daily data without considering other options. Therefore, it is necessary to analyse Bitcoin features at different frequencies. In this letter, we examine Bitcoin efficiency from 1 min to weekly data using the generalized Hurst exponent. Our results show that Bitcoin is more efficient over time regardless of the frequency. In particular, we observe that, since 2016, daily data are generally the most efficient frequency while 1 min and weekly data are the most inefficient. These results are relevant for investors and scholars since we detect the most profitable frequencies and underline the relevance of analysing different frequencies than daily data.
Bitcoin is invented in 2009 by the pseudonymous Satoshi Nakamoto. Bitcoin is a decentralized digital currency system [1]. Bitcoin is the most acknowledged cryptocurrency in the world, which provide it interesting for financier. The cryptocurrency market capitalization on date 22nd July 2020 value represents roughly USD 277 billion of dollars, bitcoin representing 62% of it. However, a disadvantage for investors is the difficulty of predicting the price of bitcoin due to the high volatility of the bitcoin exchange rate. Measurement, estimation, and modeling of currency exchange rate volatility compose a significant research area. For this reason, a lot of studies done about bitcoin price prediction both Machine Learning (ML) and Statistical Methods. In comparison studies, ML methods perform better in general. This review is a comprehensive study on how we can better predict bitcoin prices by grouping previously done studies. The presentation of Bitcoin price prediction studies in groups reveals, the difference from other review studies. These are statistical methods, ML and statistical methods, ML-ML, frequency effect of selected time, effect of social media and web search engine, causality, optimization of hyperparameters methods.
Due to its characteristics of decentralization, no counterfeit currency, and anonymity, Bitcoin has developed incredibly rapidly, gradually realizing free exchange with real currency, and stepping into the purchase of real goods and services. This paper applied historical daily frequency data of Bitcoin and constructed traditional technical indicator factors such as CCI, AROON, MA, PSY, etc. Then logistic regression and XGBoost were leveraged to predict the rise and fall of the price of Bitcoin. The results showed that XGBoost classifier obtained a higher score than two logistic regressions, therefore the XGBoost method performed better until this stage: 163% of the return, 32% of the maximum retracement. This paper helps to buy and sell Bitcoin better, get a higher positive return, and may provide ideas for the stock market research.
Edwin Mahendra, Hrithik Madan, Sonal Gupta, Sajai Vir Singh
The rapid development of digital currency especially crypto currency during the last decade is the most controversial development in the modern global economy. Because of the high volatile market and fluctuations in Bitcoin price, it has led a lot of confusion among the investors. This paper proposes the novel method of the construction of prediction model using deep learning approach. The proposed approach was found to be more accurate than the Machine learning models used for prediction as the deep learning model consider the non-linear nature of price. The results verify the applicability of model and give a direction to investors on how deep learning techniques can be used in decision making.
Bu çalışmada amaç; Bitcoin, döviz kuru, Borsa İstanbul Endeksi ve faiz değişkenleri arasındaki ilişkileri Türkiye için 2013:11-2019:10 dönemi haftalık verileri kullanarak incelemektir. Çalışmada VAR modeli kurularak değişkenler arasındaki uzun dönem ve nedensellik ilişkileri araştırılmış, etki-tepki grafikleri ve varyans ayrışım tablosuyla analiz sonuçlandırılmıştır. Çalışma sonucunda Bitcoin ile diğer değişkenler arasında uzun dönemde herhangi bir eşbütünleşme ilişkisi ve nedensellik ilişkisi tespit edilememiş, ancak diğer değişkenlerin kendi aralarında nedensellik ilişkileri saptanmıştır. Etki-tepki grafiklerine göre Bitcoin’e verilen bir şoka döviz kuru üç haftalık negatif tepki göstermiş diğer haftalarda verilen tepki anlamsız olmuştur. Türkiye’de kripto paralar üzerinde belirli bir farkındalığın olduğu ancak bu farkındalığın uzun vadeli yatırım boyutunda ve makro değişkenleri etkileyebilecek güçte olmadığı görülmektedir. Türkiye’de yeni sayılabilecek olan bu teknolojinin yaygınlaşabilmesi için belirli bir zamana ihtiyaç vardır.
The interactive information in blockchain architecture establishes an effective communication channel between users and enterprises, enabling them to communicate in a comprehensive and effective manner. Therefore, taking blockchain interactive information as the research object, this paper explores how the intervention of official information on investors affects the stock price movement and then makes predictions on stock prices according to the emotional tendency of interactive information. With the contextual information fusion, a sentiment computing model based on a convolutional neural network is established to extract and quantify the emotional features of blockchain interactive information. Combined with investors’ emotional features, the stock price prediction model based on long short-term memory is proposed. The experiment results show that the accuracy of the model has been improved by incorporating the intervened emotional features, thereby proving that information clarification can have a positive effect on the stock price.
Financial institutions such as banks regularly forecast the amount of finances an individual will have in his/her account in the near future. This can help banks in categorizing their customers so that banks can recommend financial products that matches the needs of their customers. In this work, we explored the historical financial transactions for predicting the amount a customer will receive through his/her transacting partners at a specific time. In particular, we use the Bitcoin transactional dataset, which has two main characteristics: i) network, and ii) temporal. This paper contributes by exploiting a specific kind of Graph Neural Network approach called Temporal-Graph Convolutional Network (T-GCN) for predicting the amount of Bitcoins received by a customer at a particular timestamp. The lower errors obtained using T-GCN approach compared to 11 baseline approaches (such as Support Vector Regression (SVR), Random Forest Regression (RFR), Vector Auto-Regressive (VAR), Long Short-Term Memory (LSTM), etc.) clearly demonstrate the effectiveness of T-GCN approach. In addition, our findings reveal that time is an important feature for such kind of predictive tasks.
This paper examines the sentiment spillovers among oil, gold, and Bitcoin markets by employing spillovers index methods in a time-frequency framework. We find that the total sentiment spillover among crude oil, gold and Bitcoin markets is time-varying and is greatly affected by major market events. The directional sentiment spillovers are also time-varying. On average, the Bitcoin market is the major transmitter of directional sentiment spillovers, whereas the crude oil and gold markets are the major receivers. In particular, the sentiment spillover effects are major created at high-frequency components, implying that the markets rapidly process the sentiment spillover effects and the shock is transmitted over the short-term. Moreover, we also find that the sentiment spillover effects differ significantly in term of intensity and direction when compared with return and volatility spillover effects. The present study has certain applications for investors and policymakers.
The cryptocurrency market has received immense consideration in media and academia since the beginning of 2013 because of its huge price fluctuation. This study focuses on Arab investors who invest in the cryptocurrency market by investigating the influence of behavioral finance factors on investment decisions in the cryptocurrency market. A quantitative approach was used by employing a snowball sampling method through 112 questionnaires. The results show that herding theory, prospect theory, and heuristic theory have a significant effect on investors' investment decisions in the cryptocurrency market. This emphasizes the significant role of the proposed behavioral factors as determinants of the investors' investment decisions. This study contributes to the existing research by consolidating the results of different researches in this study. It also contributes to the investors' understanding of the dynamics of the cryptocurrency market and it enhances the ability to make informed decisions based on their understanding. The implication of the findings will prepare hit and run investors to be progressively prepared to stay in the cryptocurrency market and develop their abilities on the most proficient method to settle on sound venture choices. Furthermore, the findings of this study will encourage financial specialists to realize that information on the traditional finance theory is not adequate to excel in the cryptocurrency market.
Based on the Bitcoin exchange data, COVID-19 data, and Twitter data from January 2020 to July 2020, this paper compares the performance of four different machine learning models on predicting the Bitcoin return rate and price trend. Data are formulated to four input feature sets, including: (1) Historical Bitcoin exchange data; (2) Historical Bitcoin exchange data + COVID-19 data (recovery, confirmed, death); (3) Historical Bitcoin exchange data + Twitter data; (4) Historical Bitcoin exchange data + COVID-19 data (recovery, confirmed, death) + Twitter data. The four machine learning models implemented are: (1) Random forest; (2) Decision tree; (3) AdaBoost; (4) Support vector machine. We found that: (1) Twitter data can improve the performance of models; (2) People consider information within 5 days when they make decisions on investments; (3) Support vector machine does not perform well in predicting Bitcoin return rate or price trend; (4) COVID-19 data does not help improve the prediction. However, we have very limited COVID-19 data, so future research with more COVID-19 data may help confirm if the last statement is correct or not.
COVID-19 affects significant human activity around the globe, including Bitcoin prices. The Bitcoin price is well known for its volatility, so it is not a big shocker when the panic-selling occurs during the pandemic. However, the mechanism to cope with these breakouts, especially the bearish one, is contentious. The experts give numerous pieces of advice with different conclusions in the end. It is also the same with Machine Learning. Various kernels show different results regarding how the price will move. It depends on the window size, how the data is being preprocessed, and the algorithm used. This paper inspects the best combination that various machine learning can offer with a linear approach to navigate the price prediction based on its depth interval, window size until the algorithms themselves. This paper also proposed a new approach to seeing the prediction range called s-steps ahead prediction using a linear model. The result shows that simple machine learning can herd 99.715% profit even during the bearish breakout.
Bitcoin is the most popular and valuable cryptocurrency in the financial market which attracts traders for investment and opens new research opportunities for researchers. Many research works have been done on bitcoin price prediction with different machine learning prediction algorithms. Researchers take relevant features from the dataset which have strong correalation with bitcoin prices and select random data chunks to train and test their model. Randomly selected data to train the model, may cause inappropriate results and reduce the accuracy of price prediction. In this paper, we investigate a proper data selection method to train a prediction model. We apply our proposed methodology to train a simple linear regression prediction algorithm. We predict bitcoin price for 7 days with the linear regression model. When we train the linear regression model with an appropriate data chunk identified by our methodologies, we find acceptable results for the prediction. The percentage error method is applied for error calculation which finds the accuracy is 96.97%. In the end of this manuscript, we conclude our work with future improvements.
Salim Lahmiri, Raafat George Saadé, Danielle Morin, Fassil Nebebe
Cryptocurrencies are digital assets gaining popularity and generating huge transactions on electronic platforms. We develop an ensemble predictive system based on artificial neural networks to forecast Bitcoin daily trading volume level. Indeed, although ensemble forecasts are increasingly employed in various forecasting tasks, developing an intelligent predictive system for Bitcoin trading volume based on ensemble forecasts has not been addressed yet. Ensemble Bitcoin trading volume are forecasted using two specific artificial neural networks; namely, radial basis function neural networks (RBFNN) and generalized regression neural networks (GRNN). They are adopted to respectively capture local and general patterns in Bitcoin trading volume data. Finally, the feedforward artificial neural network (FFNN) is implemented to generate Bitcoin final trading volume after having aggregated the forecasts from RBFNN and GRNN. In this regard, FFNN is executed to merge local and global forecasts in a nonlinear framework. Overall, our proposed ensemble predictive system reduced the forecasting errors by 18.81% and 62.86% when compared to its components RBFNN and GRNN, respectively. In addition, the ensemble system reduced the forecasting error by 90.49% when compared to a single FFNN used as a basic reference model. Thus, the empirical outcomes show that our proposed ensemble predictive model allows achieving an improvement in terms of forecasting. Regarding the practical results of this work, while being fast, applying the artificial neural networks to develop an ensemble predictive system to forecast Bitcoin daily trading volume is recommended to apply for addressing simultaneously local and global patterns used to characterize Bitcoin trading data. We conclude that the proposed artificial neural networks ensemble forecasting model is easy to implement and efficient for Bitcoin daily volume forecasting.
Many traders participate in activities known as "day-trading", trading Bitcoin against the dollar bill as the United States Dollar (USD) on very short timeframes to squeeze out profits from small market fluctuations. This paper aims to help traders decide how to best act by creating a model that can predict price movement's direction for the next 5-min time frame. Several machine-learning models have been tested for this Up/Down binary-classification problem. In this paper, we provide a comparison of the state-of-art strategies in predicting the movement direction for bitcoin, including Random Guessing and a Momentum-Based Strategy. The tested models include Autoregressive Integrated Moving Average (ARIMA), Prophet (by Facebook), Random Forest, Random Forest Lagged-Auto-Regression, and Multi-Layer Perceptron (MLP) Neural Networks. The MLP deep neural network has achieved the highest accuracy of 54% compared to other time-series prediction models. Also, in this paper, various data transformation and feature engineering have been applied in the comparison.