The issue of market efficiency for cryptocurrency exchanges has been largely unexplored. Here we put Bitcoin, the leading cryptocurrency, on a test by studying the applicability of the Efficient Market Hypothesis by Fama from two viewpoints: (1) the existence of profitable arbitrage spread among Bitcoin exchanges, and (2) the possibility to predict Bitcoin prices in EUR (time period 2013-2017) and the direction of price movement (up or down) on the daily trading scale. Our results show that the Bitcoin market in the time period studied is partially inefficient. Thus the market process is predictable to a degree, hence not a pure martingale. In particular, the F-measure for XBTEUR time series obtained by three major recurrent neural network based machine learning methods was about 67%, i.e. a way above the unbiased coin tossing odds of 50% equal chance.
Bitcoin it is digital currency used for any Electronic Financial Transactions, created in 2008 by Satoshi Nakamoto as Peer-to-Peer Electronic System cash. The precise prediction of bitcoin exchange rate with respect to the US dollar is an essential matter because it effected on the world economic constancy. Meanwhile the technology that used in Bitcoin has flickered a revolution in world. This paper aim to a survy on the use artificial neural networks to predict the exchange rate of a Bitcoin(BTC ) currency. https://doi.org/10.24897/acn.64.68.172
This paper is a sequel to our previous works related to Robo-advisors and cryptocurrencies. Our goal now is to build two application modules for a single Robo-advisor. The first module is a Long short-term memory (LSTM) neural network which forecasts cryptocurrencies prices daily. The second module uses Robo-advising approach to build an investment plan for novice cryptocurrencies investors with different risk attitude investment decisions. The third module does ETL (Extract-Transform-Load) for a statistics dataset and neural networks models. Results of the investigation show that investing in cryptocurrencies can give 23.7% per year for risk-averse, 31.8% per year for risk-seeking investors and 16.5% annually for riskneutral investors.
Machine Learning is part of Artificial Intelligence that has the ability to make future forecastings based on the previous experience. Methods has been proposed to construct models including machine learning algorithms such as Neural Networks (NN), Support Vector Machines (SVM) and Deep Learning. This paper presents a comparative performance of Machine Learning algorithms for cryptocurrency forecasting. Specifically, this paper concentrates on forecasting of time series data. SVM has several advantages over the other models in forecasting, and previous research revealed that SVM provides a result that is almost or close to actual result yet also improve the accuracy of the result itself. However, recent research has showed that due to small range of samples and data manipulation by inadequate evidence and professional analyzers, overall status and accuracy rate of the forecasting needs to be improved in further studies. Thus, advanced research on the accuracy rate of the forecasted price has to be done.
This paper presents all studies, methodology, and results about Bitcoin forecasting with PROPHET and ARIMA methods using R analytics platform. To find the most accurate forecast model, the performance metrics of PROPHET and ARIMA methods are compared on the same dataset. The dataset selected for this study starts from May 2016 and ends in March 2018, which is the interval that Bitcoin values changing significantly against the other currencies. Data is prepared for time series analysis by performing data preprocessing steps such as time stamp conversion and feature selection. Although the time series analysis has a univariate characteristics, it is aimed to include some additional variables to each model to improve the forecasting accuracy. Those additional variables are selected based on different correlation studies between cryptocurrencies and real currencies. The model selection for both ARIMA and PROPHET is done by using threefold splitting technique considering the time series characteristics of the dataset. The threefold splitting technique gave the optimum ratios for training, validation, and test sets. Finally two different models are created and compared in terms of performance metrics. Based on the extensive testing we see that PROPHET outperforms ARIMA by 0.94 to 0.68 in R2values.
The decentralization of cryptocurrencies has greatly reduced the level of central control over them, impacting international relations and trade. Further, wide fluctuations in cryptocurrency price indicate an urgent need for an accurate way to forecast this price. This paper proposes a novel method to predict cryptocurrency price by considering various factors such as market cap, volume, circulating supply, and maximum supply based on deep learning techniques such as the recurrent neural network (RNN) and the long short-term memory (LSTM),which are effective learning models for training data, with the LSTM being better at recognizing longer-term associations. The proposed approach is implemented in Python and validated for benchmark datasets. The results verify the applicability of the proposed approach for the accurate prediction of cryptocurrency price.
Pavitra Mohanty, Darshan Patel, Parth Patel, Sudipta Roy
This paper shows the prediction of fluctuation in the future price of cryptocurrencies. Users' comments and tweets from twitter using Apache Flume and Price data was fetched from exchanges. Bitcoin first documented by allies Satoshi Nakamoto, the first decentralized currency payment system has gained a considerable attention in the financial system, economics, social media and computer science due to its combination of peer-to-peer nature, encryption technology, and monetary unit. Predicting the price of Bitcoin and other cryptocurrencies is a great challenge because it is immensely complex and dynamic in nature. In this paper, we have tried to predict the future price of cryptocurrencies like Bitcoin using LSTM (Long Short-Term Memory) and used Twitter data to predict public mood. By combining both market sentiment and social sentiment because bitcoin price shows mixed properties. We also have selected some other important features from the blockchain information which has a major impact on Bitcoin's supply and demand and using them to train model that improves the predictive power of the future Bitcoin price. We have performed a deep study of how data from social media affect the price of Bitcoin and so we have included the twitter data in model training. Our model shows that how well LSTM predict the price of Bitcoin considering the high volatility. The precision given by our model is 60% and accuracy is 50%. More focus is not given to accuracy, in this case, considering the highly volatile market.
Crypto currencies are considered as the next model of economics and monetary exchange. In recent years, popular cryptocurrency such as Bitcoin and Ethereum witness an exponential growth in economic sphere. In this paper empirical testing of four conventional machine learning methods is performed to predict the bitcoin prices using last eight years of transactional data. Linear and polynomial regression is implemented using all the features individually. Polynomial regression, Support Vector regression and KNN regression are hyper tuned with grid search logic. Results depicted that KNN regression outperformed others models in attaining mean square error of 0.00021.
Arti Jain, Shashank Tripathi, Harsh Dhar Dwivedi, Pranav Saxena
The problem is to find a method to predict the two-hour price of cryptocurrencies on the basis of the Social Factors, which are increasingly used for online transactions worldwide. The few previous methods proposed to predict price of cryptocurrency are inefficient because they fail to take into consideration the differences in the attributes between real currencies and cryptocurrencies. In this paper, we focus on two cryptocurrencies, namely Bitcoin and Litecoin, each with a large market size and user base, and attempt to predict their future prices using multi-linear regression model.
Kripto para birimleri teknolojinin gelişmesiyle birlikte son yıllarda önem kazanmış ve daha çok kullanılır hale gelmiştir. Merkezi bir otoriteye bağlı olmayan ve kriptografik sistemler ile güvenliği sağlanan bu para birimlerinden en bilineni Bitcoin’dir. Bu çalışmada, başlıca kripto para birimleri ve işleyiş süreçleri incelenmiştir. Buna ek olarak Bitcoin’in döviz, hisse senedi emtia piyasaları ve faiz ile olan ilişkisi ele alınmıştır. Çalışmada kullanılan veri setinin frekansı aylık olup Mart-2012 ile Mayıs-2018 dönemini kapsamaktadır. Zaman serisi yöntemlerinden Johansen Eşbütünleşme ve Granger Nedensellik analizleri uygulanmıştır. Çalışmanın sonuçlarına göre, Bitcoin fiyatlarının artan bir trende ve yüksek bir volatiliteye sahip olduğu görülmektedir. Faiz değişkeni ile Bitcoin fiyatları arasında diğer analizler ve Granger nedensellik testi sonuçlarına göre istatistiksel olarak anlamlı bir ilişki vardır.
For various reasons, financial institutions often make use of high-level trading strategies when buying and selling assets. Many individuals, irrespective or their level of prior trading knowledge, have recently entered the field of trading due to the increasing popularity of cryptocurrencies, which offer a low entry barrier for trading. Regardless of the intention or trading strategy of these traders, the invariable outcome is their attempt to buy or sell assets. However, in such a competitive field, experienced market participants seek to exploit any advantage over those who are less experienced, for financial gain. Therefore, this work aims to make a contribution to the important issue of how to optimize the process of buying and selling assets on exchanges, and to do so in a form that is accessible to other traders. This research concerns the optimization of limit order placement within a given time horizon of 100 seconds and how to transpose this process into an end-to-end learning pipeline in the context of reinforcement learning.<br/>Features were constructed from raw market event data that related to movements of the Bitcoin/USD trading pair on the Bittrex cryptocurrency exchange. These features were then used by deep reinforcement learning agents in order to learn a limit order placement policy. To facilitate the implementation of this process, a reinforcement learning environment that emulates a local broker was developed as part of this work. Furthermore, we defined an evaluation procedure which can determine the capabilities and limitations of the policies learned by the reinforcement learning agents and ultimately provides means to quantify the optimization achieved with our approach. Our analysis of the results of this work includes the identification of patterns in cryptocurrency trading that were formed by market participants who posted orders, and a conceptual framework to construct data features containing these patterns. We developed a fully-functioning reinforcement learning environment that emulates a local broker and, by means of this process, we identified which components are essential.<br/>With the use of this environment, we were able to train and test multiple reinforcement learning agents whose aims were to optimize the placement of buy and sell limit orders. During the evaluation, we were able to improve the parameter settings of the constructed reinforcement learning environment and therefore improve the policy learned by the agents. Ultimately, we achieved a significant improvement in limit order placement with the application of a state-of-the-art deep Q-network agent and were able to simulate purchases and sales of 1.0 BTC at a price that was up to $33.89 better than the market price. We have made use of the OpenAI Gym library and contributed our work to the community to enable further investigations to be carried out. The work done in this thesis can be used as a framework to (1) build a component that acts as an intermediary between trader and exchange and (2) to enable exchanges to provide a new order type to be used by traders.
(eng) The main conceptual element this thesis orbits around is the idea of using social networks as a data source. First, classical trading theory and current usage of data obtained from social networks is reviewed. Taking all this information into account, a forecasting of the Bitcoin price is performed using both classical methods and machine learning Neural Networks. In order to obtain data from social networks, another complexity layer needs to be added by accessing the sources through APIs and directly web-scrapping the net. The results of all of this complex implementation are given with a strong focus on visualisation using several different techniques. Finally, after a critical discussion a Future Work chapter is introduced, where many possible follow-ups are drawn up.
This paper explores Bitcoin intraday technical trading based on artificial neural networks for the return prediction. In particular, our deep learning method successfully discovers trading signals through a seven layered neural network structure for given input data of technical indicators, which are calculated by the past time-series data over every 15 minutes. Under feasible settings of execution costs, the numerical experiments demonstrate that our approach significantly improves the performance of a buy-and-hold strategy. Especially, our model performs well for a challenging period from December 2017 to January 2018, during which Bitcoin suffers from substantial minus returns. Furthermore, various sensitivity analysis is implemented for the change of the number of layers, activation functions, input data and output classification to confirm the robustness of our approach.
Taiguara Melo Tupinambás, Rafael Aeraf Leao Cadence, André Lemos
Cryptocurrencies prices forecasting is a complex theme due to the chaotic market behavior and the influence of external events. Therefore, inference models should offer, in addition to a satisfying accuracy, reasonable interpretability, so that investors can decide based on their own knowledge. However, many studies in this subject focus on model accuracy and leave much to be desired in terms of simplicity and interpretability. This work proposes the use of Mamdani interpretable fuzzy inference models for forecasting cryptocurrency price variation. For that, a genetic algorithm to optimize models accuracy is employed, limiting the quantity of rules and antecedents arbitrarily. A set of infeasible rules had to be discarded, in order to generate interesting models, that produce a relevant amount of trades. Data from Kraken exchange were utilized for training, validation and results assessment. Results have shown that, for the cryptocurrencies with the highest validation performances, there are gains in comparison to the simple currency appreciation. Using the interpretable aspect of the models, it should be possible to obtain even higher profits.
Currently, Cryptocurrency is one of the trending areas of research among researchers. Many researchers may analyze the cryptocurrency features in several ways such as market price prediction, the impact of cryptocurrency in real life and so on. In this paper, we focus on market price prediction of the number of cryptocurrencies based on their historical trend. For our study, we tried to understand and identify the daily trends in the cryptocurrency market which analyzing the features related to the price of cryptocurrency. Our dataset consists of over nine features relating to the cryptocurrency price recorded daily over the period of 6 months. We applied some machine-learning algorithms to predict the daily price change of cryptocurrencies.
In recent years, Bitcoin is the most valuable in the cryptocurrency market. However, prices of Bitcoin have highly fluctuated which make them very difficult to predict. Hence, this research aims to discover the most efficient and highest accuracy model to predict Bitcoin prices from various machine learning algorithms. By using 1-minute interval trading data on the Bitcoin exchange website named bitstamp from January 1, 2012 to January 8, 2018, some different regression models with scikit-Iearn and Keras libraries had experimented. The best results showed that the Mean Squared Error (MSE) was as low as 0.00002 and the R-Square (R2) was as high as 99.2%.
Kripto parateknolojisi, insanların herhangi bir aracıya ihtiyaç duymadan dünyanın dört biryanında ödeme yapmalarını sağlar ve internet üzerinden çalışır. Bu teknolojininortaya çıkışı ile dikkatler bu paralara çevrilmiş, fiyatları hızla artmayabaşlamış ve aynı zamanda oynak hale gelmişlerdir. Bu çalışmada GARCH (1,1)modeli kullanılarak Bitcoin ve Litecoin piyasalarında haftanın günü ve yılınayı etkilerinin varlığı incelenmiştir. Çalışma dönemi 1 Mayıs 2013-21 Aralık2016 arasını kapsamaktadır. Elde edilen sonuçlar Bitcoin ve Litecoingetirilerinde haftanın günü ve yılın ayı etkilerinin var olduğunugöstermektedir. Pazartesi, Salı ve Cuma günlerinin Bitcoin getirileri üzerindepozitif ve anlamlı, Cumartesi’nin ise Litecoin getirileri üzerinde negatif veanlamlı etkisi olduğu tespit edilmiştir. Ayrıca, yılın ayı etkisi açısındanŞubat, Ekim ve Kasım aylarının Bitcoin üzerinde pozitif ve anlamlı, Litecoingetirilerinde ise Ağustos ayının negatif ve anlamlı etkisiolduğubelirlenmiştir.
Die Preisvorhersage ist eine der größten Herausforderungen der quantitativen Finanzierung. Diese Thesis stellt ein Neuronales Netz-Framework vor, das eine tiefgreifende maschinelle Lernlösung für das Preisvorhersageproblem bietet. Das Framework wird in drei Zeitpunkten mit einem Multilayer Perzeptron (MLP), einem einfachen Recurrent Neural Network (RNN) und einem Long Short Term Memory (LSTM) realisiert, die lange Abhängigkeiten lernen können. Wir beschreiben die Theorie der neuronalen Netze und des Deep Learning, um eine reproduzierbare Methode für unsere Anwendungen auf dem Kryptowährungsmarkt zu erstellen. Da die Preisvorhersage verwendet wird, um finanzielle Entscheidungen wie Handelssignale zu treffen, vergleichen wir verschiedene Ansätze des Vorhersageproblems, indem wir überwachte Lernmethoden in Klassifikationsaufgaben untersuchen. Wir untersuchen diese Modelle, um Preisrichtungen von wichtitgen Kryptowährungen außerhalb der Stichprobe mit einer rolling window regression Methode vorherzusagen. Für dieses Ziel erstellen wir ein Klassifikationsproblem, das voraussagt, ob der Preis jeder Kryptowährung als Grundlage für dreimonatige Handelsstrategien erheblich zu- oder abnimmt. Wir bauen verschiedene Handelsstrategien auf, basierend auf Long- oder Long- / Short-Positionen, die auf unseren Prognosen aufbauen, und vergleichen ihre Performance mit einer passiven Index-Investition auf dem Cryptowährungsmarkt, die CRIX (Trimborn and Härdle, 2016) folgt. Cryptocurrencies, Bitcoins sind die bekanntesten, basieren auf elektronischem Geld auf Blockchain-Technologie, die als eine dezentrale Alternative zu Währungen verwendet werden kann. Dank ihrer zahlreichen Anwendungen hat der Markt für Kryptowährung im Jahr 2017 ein exponentielles Wachstum erfahren. Wir vergleichen verschiedene gewichtete Portfolios, um zu testen, wie ein Anleger von fundamentalen Indikatoren wie der Marktkapitalisierung profitieren kann. Wir finden, dass LST die beste Genauigkeit für die Vorhersage von Richtungsbewegungen für die wichtigsten Kryptowährungen von CRIX hat und dass ein gleich gewichtetes Portfolio CRIX in den ersten Quartalen 2017 schlägt.
Bu çalışmada her geçen gün ilgiyle izlenmeye devam edilen sanal para birimi Bitcoin’de çoklu balonların varlığı Phillips, Shi ve Yu (2015) tarafından geliştirilen GSADF birim kök testi ve kritik değerlerin tespitinde her türlü değişen varyans problemini hesaba katarak işlem yapan Harvey, Leybourne, Sollis ve Taylor (2016) tarafından geliştirilen metot takip edilerek araştırılmıştır. Veri seti 16.07.2010 ve 31.12.2017 tarihleri arasında günlük bazdaki 24 saatlik ortalama Bitcoin fiyatlarından oluşmaktadır. Yapılan analizler sonucunda söz konusu veri aralığının büyük bir kısmında Bitcoin fiyatlarında çoklu balonların varlığı görülmüştür
Jun 1, 2018·2018 Joint 7th International Conference on Informatics, Electronics & Vision (ICIEV) and 2018 2nd International Conference on Imaging, Vision & Pattern Recognition (icIVPR)
Shaomi Rahman, Jonayed Nafis Hemel, Syed Junayed Ahmed Anta, Hossain Al Muhee · 5 authors
In this paper, we have proposed the correlation between the price change of Bitcoin and its user's sentiment by implementing machine learning algorithms. In this model, we have clearly described our goals of implementation, the process of implementation along with its final analysis and the considering predicted price change and actual price change. We approached the ambitious problem of predicting Bitcoin price change with sentiment in the hope that we find the significance of people's opinion in the field of cryptocurrency. Also, this research introduces a new way of utilizing social networking sites' data.
The purpose of this study is to apply the α-Sutte Indicator and ARIMA in forecasting data. α-Sutte Indicator is a new forecasting method that was developed in 2017 by Ansari Saleh Ahmar. To see the accuracy of these methods, the forecasting results of the α-Sutte Indicator will be forecasting methods compared to other items, namely: ARIMA. Based on the results of forecasting, it is found that α-Sutte Indicator has MSE and MAE values that are lower than other methods (ARIMA). This is supported by MSE data from α-Sutte Indicator smaller than ARIMA(1,1,1).
In the world finance and technological development in finance, along with innovative financial instruments, have attracted investors. The most popular of these developments is undoubtedly Bitcoin, which is an output of the blockchain infrastructure .Bitcoin that is not connected to a central authority and contains cryptographic features, is one of the crypto moneys. The fact that Bitcoin does not depend on Central Authority and disclose the factors affecting its price by supply and demand have resulted in high volatility. In this study, firstly blockchain technology will be explained briefly and time-dependent price estimates for Bitcoin which is one of the important outputs of this technology, will be made. Artificial Neural Networks (YSA), which has become increasingly popular among estimation methods in recent years, has been used in the study and compared with ARIMA in traditional estimation methods. The sample of the study was created using daily closing prices between 02.02.2012 - 09.01.2018 dates. As a result of this study, both directions and values of estimated prices by artificial neural networks MPL (6-3-1) model between 10.01.2018 - 18.01.2018 have been more successful than ARIMA (1.1.6) model.
In recent years bitcoin has been attracted eminent attraction all around world in the area of cryptocurrency and its unique peer to peer transaction system. Analyzing and forecasting is most common task for data scientists that helps organizations to improve there business strategies. \n \nThis report is built around the fact that in bitcoin trading and analysis, the factors underlying are basically the traditional price predictions using the data mining technique.The project aim to gather data on bitcoin, analyse and forecast in order to see the fall or rise in bitcoin price.The bitcoin Dataset used is bitcoinmarketcap. This report intend to help the reader and understand what approach and methodology used to complete different milestones.This report summarizing all the aspect used throughout this project. \n \nThis is done using different techniques and algorithm to predict and forecast the rise or fall of bitcoin with the use of Crisp-DM methodology.The techniques used in the projects are Time Series ARIMA ,Time Series( Facebook prophet API), Linear Regression and and Multiple Regression.