<title>Abstract</title> <bold>Background: </bold>Cryptocurrencies, especially Bitcoin, has become popular for investors in recent years. The volatility of bitcoin and time horizon are the center point for investment decisions. However, attention is not often drawn to the relationship between bitcoin and equity indices. This study investigates the volatility and time frequency domain of bitcoin among five Asean countries through a rich database which covers daily data from July 2010 until April 2019.<bold>Methods: </bold>Advanced econometrics and Wavelets Cross-Coherence Spectrograms, this study investigates the existence of long run association between bitcoin and the studied market indices. M-GARCH analysis is been employed to investigate the unconditional volatility of market indices and Bitcoin.<bold>Results: </bold>The findings present the long run association<bold> </bold>with positive (Philippines) and negative (Japan, Korea, Singapore, Hong Kong) relations. Moreover, only one market (KOREA) shows a short run association with bitcoin. The M-GARCH analysis reveals, most of the selected Asean countries have a low unconditional volatility with bitcoin. Except for Philippines in which the co-movement is average, Wavelet analysis reveals the presence of a strong and long co-movements for most of the selected Asean countries with bitcoin.<bold>Conclusions: </bold>Most of our results are consistent and illustrate different dimensions of long and short run relationship, volatilities, correlations, and time-frequency analysis. This study utilized Asean emerging economies which are rarely available in the literature as existing studies are more skewed towards the West. We believe the outcomes of this study will be a significant for industry practitioners (i.e., retail and institutional investors) on designing better strategies to diversify the stock portfolio with different holding period horizons and dimensions.
Ioannis E. Livieris, Emmanuel Pintelas, Stavros Stavroyiannis, Panagiotis Pintelas
Nowadays, cryptocurrency has infiltrated almost all financial transactions; thus, it is generally recognized as an alternative method for paying and exchanging currency. Cryptocurrency trade constitutes a constantly increasing financial market and a promising type of profitable investment; however, it is characterized by high volatility and strong fluctuations of prices over time. Therefore, the development of an intelligent forecasting model is considered essential for portfolio optimization and decision making. The main contribution of this research is the combination of three of the most widely employed ensemble learning strategies: ensemble-averaging, bagging and stacking with advanced deep learning models for forecasting major cryptocurrency hourly prices. The proposed ensemble models were evaluated utilizing state-of-the-art deep learning models as component learners, which were comprised by combinations of long short-term memory (LSTM), Bi-directional LSTM and convolutional layers. The ensemble models were evaluated on prediction of the cryptocurrency price on the following hour (regression) and also on the prediction if the price on the following hour will increase or decrease with respect to the current price (classification). Additionally, the reliability of each forecasting model and the efficiency of its predictions is evaluated by examining for autocorrelation of the errors. Our detailed experimental analysis indicates that ensemble learning and deep learning can be efficiently beneficial to each other, for developing strong, stable, and reliable forecasting models.
Clément Bénard, Gérard Biau, Sébastien da Veiga, Erwan Scornet
We introduce SIRUS (Stable and Interpretable RUle Set) for regression, a stable rule learning algorithm which takes the form of a short and simple list of rules. State-of-the-art learning algorithms are often referred to as "black boxes" because of the high number of operations involved in their prediction process. Despite their powerful predictivity, this lack of interpretability may be highly restrictive for applications with critical decisions at stake. On the other hand, algorithms with a simple structure-typically decision trees, rule algorithms, or sparse linear models-are well known for their instability. This undesirable feature makes the conclusions of the data analysis unreliable and turns out to be a strong operational limitation. This motivates the design of SIRUS, which combines a simple structure with a remarkable stable behavior when data is perturbed. The algorithm is based on random forests, the predictive accuracy of which is preserved. We demonstrate the efficiency of the method both empirically (through experiments) and theoretically (with the proof of its asymptotic stability). Our R/C++ software implementation sirus is available from CRAN.
Petrônio Cândido de Lima e Silva, Carlos Alberto Severiano, Marcos Antônio Alves, Rodrigo Silva · 6 authors
In this paper we introduce a Non-Stationary Fuzzy Time Series (NSFTS) method with time varying parameters adapted from the distribution of the data. In this approach, we employ Non-Stationary Fuzzy Sets, in which perturbation functions are used to adapt the membership function parameters in the knowledge base in response to statistical changes in the time series. The proposed method is capable of dynamically adapting its fuzzy sets to reflect the changes in the stochastic process based on the residual errors, without the need to retraining the model. This method can handle non-stationary and heteroskedastic data as well as scenarios with concept-drift. The proposed approach allows the model to be trained only once and remain useful long after while keeping reasonable accuracy. The flexibility of the method by means of computational experiments was tested with eight synthetic non-stationary time series data with several kinds of concept drifts, four real market indices (Dow Jones, NASDAQ, SP500 and TAIEX), three real FOREX pairs (EUR-USD, EUR-GBP, GBP-USD), and two real cryptocoins exchange rates (Bitcoin-USD and Ethereum-USD). As competitor models the Time Variant fuzzy time series and the Incremental Ensemble were used, these are two of the major approaches for handling non-stationary data sets. Non-parametric tests are employed to check the significance of the results. The proposed method shows resilience to concept drift, by adapting parameters of the model, while preserving the symbolic structure of the knowledge base.
Nowadays, accurate prediction of cryptocurrency price variation based on their important role in the world economy is an important and challenging issue. In this study, various parameters that affect the cryptocurrency value have been considered. For the first phase, four major price features of digital currencies have been analysed to determine the effect of each feature on the volatility prediction of future days. This study aims to understand and identify daily trends in the cryptocurrency market while gaining insight into optimal features surrounding their price. For the second phase, the price variation has been predicted with the highest possible accuracy with a new intelligent method. The proposed method consists of a neural network‐based prediction algorithm and particle swarm optimisation. The obtained results show the capbility of the proposed method.
Roy Cerqueti, Massimiliano Giacalone, Raffaele Mattera
Recently, cryptocurrencies have attracted a growing interest from investors,\npractitioners and researchers. Nevertheless, few studies have focused on the\npredictability of them. In this paper we propose a new and comprehensive study\nabout cryptocurrency market, evaluating the forecasting performance for three\nof the most important cryptocurrencies (Bitcoin, Ethereum and Litecoin) in\nterms of market capitalization. At this aim, we consider non-Gaussian GARCH\nvolatility models, which form a class of stochastic recursive systems commonly\nadopted for financial predictions. Results show that the best specification and\nforecasting accuracy are achieved under the Skewed Generalized Error\nDistribution when Bitcoin/USD and Litecoin/USD exchange rates are considered,\nwhile the best performances are obtained for skewed Distribution in the case of\nEthereum/USD exchange rate. The obtain findings state the effectiveness -- in\nterms of prediction performance -- of relaxing the normality assumption and\nconsidering skewed distributions.\n
Abstract— After being introduced in 2008, the rise in the price of bitcoin and the popularity of other cryptocurrencies triggered a growing discussion about how much energy was consumed during the production of this currency. Making cryptocurrency the most expensive and most popular, both the business world and the research community have begun to study the devel-opment of bitcoin. In this study bitcoin price predictions are performed using the double exponential smoothing method based on the mean absolute percentage error (MAPE). The MAPE value is used to find the best alpha (α) parameter as the basis for bitcoin price forecasting. The dataset used is the price of bitcoin from 2017 to 2019. The dataset was obtained from www.cryptocompare.com. As for the value of the alpha parameter (α), using a value of 0.1 to 0.9. Based on the test results using the double exponential smoothing method obtained the smallest MAPE value of 2.89%, with the best alpha (α) at 0.9. The prediction is done to see the price of bitcoin on January 1, 2020. The error rate generated on the predicted price of bitcoin uses an amount of 0.0373%. This shows that the system built can be used as a support for decision making when trading bitcoin.
Blockchain teknolojisinin aracısız veri/para transferi gerçekleştirmesi ile tüm çevrelerin gündemine gelen Bitcoin, birçok yatırımcının ilgi odağı olmuş ve Bitcoin ’in popüler olması ile birçok kripto para piyasaya sürülmüştür. Kripto paralarda fiyat volatilitesinin yüksekliği hızlı para kazanma arzusu içinde olan ve risk iştahı yüksek olan yatırımcıları fiyat tahminlemesi ve fiyatları etkileyen değişkenlerin belirlenmesi noktasında analiz yapmaya itmiştir. Bu çalışmanın amacı Aralık 2017 itibari ile değeri yaklaşık 20.000 ABD Dolara ulaşan ve yüksek volatilitesi ile yatırımcıların sürekli gündeminde olan Bitcoin fiyatına etki eden faktörlerin belirlenmesidir. Bu amaçla çalışmada literatürde kullanılan değişkenlere (Altın ve ABD Dolar) ek olarak küresel risklerin (Finansal Baskı Endeksi ve Jeopolitik Risk Endeksi) etkisi de ölçülmeye çalışılmıştır. Çalışama da Bitcoin fiyatı üzerine etki etmesi muhtemel değişkenler Çok Değişkenli Uyarlanabilir Regresyon Uzanımları-MARS yöntemi ile analiz edilmiştir. Çalışmada kullanılan veriler 2012/1-2019/11 yılları arasında aylık verilerden oluşmaktadır. Çalışmanın sonucunda kullanılan tüm bağımsız değişkenlerin belirli şartlar altında Bitcoin fiyatına etki edebileceği sonucuna ulaşılmıştır.
There has been a recent surge in interest in the application of artificial intelligence to automated trading. Reinforcement learning has been applied to single- and multi-instrument use cases, such as market making or portfolio management. This paper proposes a new approach to framing cryptocurrency market making as a reinforcement learning challenge by introducing an event-based environment wherein an event is defined as a change in price greater or less than a given threshold, as opposed to by tick or time-based events (e.g., every minute, hour, day, etc.). Two policy-based agents are trained to learn a market making trading strategy using eight days of training data and evaluate their performance using 30 days of testing data. Limit order book data recorded from Bitmex exchange is used to validate this approach, which demonstrates improved profit and stability compared to a time-based approach for both agents when using a simple multi-layer perceptron neural network for function approximation and seven different reward functions.
Eng Chuen Loh, Shuhaida Ismail, Azme Khamis, Aida Mustapha
Bitcoin is the most popular cryptocurrency with the highest market value. It was said to have potential in changing the way of trading in future. However, Bitcoin price prediction is a hard task and difficult for investors to make decision. This is caused by nonlinearity property of the Bitcoin price. Hence, a better forecasting method are essential to minimize the risk from inaccuracy decision. The aim of this paper is to compare two different training algorithms which are Levenberg-Marquardt (LM) backpropagation algorithm and Scaled Conjugate Gradient (SCG) backpropagation algorithm using Feedforward Neural Network (FNN) to forecast the Bitcoin price. After obtaining the forecasting result, forecast accuracy measurement will be carried out to identify the best model to forecast Bitcoin price. The result showed that the performance of Bitcoin price forecasting increased after the application of FNN – LM model. It is proven that Levenberg-Marquardt backpropagation algorithm is better compared to Scaled Conjugate Gradient backpropagation when forecasting Bitcoin price using FNN. The resulting model provides new insights into Bitcoin forecasting using FNN – LM model which directly benefits the investors and economists in lowering the risk of making wrong decision when it comes to invest in Bitcoin. Keywords: Bitcoin Price; Artificial Neural Network; Forecasting
With the proliferation of blockchain projects and applications, cryptocurrency exchanges, which provides exchange services among different types of cryptocurrencies, become pivotal platforms that allow customers to trade digital assets on different blockchains. Because of the anonymity and trustlessness nature of cryptocurrency, one major challenge of crypto-exchanges is asset safety, and all-time amount hacked from crypto-exchanges until 2018 is over $1.5 billion even with carefully maintained secure trading systems. The most critical vulnerability of crypto-exchanges is from the so-called hot wallet, which is used to store a certain portion of the total asset online of an exchange and programmatically sign transactions when a withdraw happens. It is important to develop network security mechanisms. However, the fact is that there is no guarantee that the system can defend all attacks. Thus, accurately controlling the available assets in the hot wallets becomes the key to minimize the risk of running an exchange. In this paper, we propose Shoreline, a deep learning-based threshold estimation framework that estimates the optimal threshold of hot wallets from historical wallet activities and dynamic trading networks.
With the proliferation of blockchain projects and applications, cryptocurrency exchanges, which provides exchange services among different types of cryptocurrencies, become pivotal platforms that allow customers to trade digital assets on different blockchains. Because of the anonymity and trustlessness nature of cryptocurrency, one major challenge of crypto-exchanges is asset safety, and all-time amount hacked from crypto-exchanges until 2018 is over $1.5 billion even with carefully maintained secure trading systems. The most critical vulnerability of crypto-exchanges is from the so-called hot wallet, which is used to store a certain portion of the total asset of an exchange and programmatically sign transactions when a withdraw happens. Whenever hackers managed to gain control over the computing infrastructure of the exchange, they usually immediately obtain all the assets in the hot wallet. It is important to develop network security mechanisms. However, the fact is that there is no guarantee that the system can defend all attacks. Thus, accurately controlling the available assets in the hot wallets becomes the key to minimize the risk of running an exchange. However, determining such optimal threshold remains a challenging task because of the complicated dynamics inside exchanges. In this paper, we propose Shoreline, a deep learning-based threshold estimation framework that estimates the optimal threshold of hot wallets from historical wallet activities and dynamic trading networks. We conduct extensive empirical studies on the real trading data from a trading platform and demonstrate the effectiveness of the proposed approach.
This study explores the portfolio management of cryptocurrencies by assessing the out-of-sample performance of selected portfolio strategies in the literature. Using daily data from 500 randomly selected cryptocurrencies with monthly and weekly revision, the scaled and stable mean-variance-entropic (MVE) value-at-risk portfolios outperform other portfolio strategies closely followed by 1/N portfolios. The mean Sharpe ratio with transaction costs of both MVE and 1/N was higher than that of benchmark, Coinbase index. Indeed, diversification across cryptocurrencies does improve investment results and mitigates risk exposure. The findings of this research are crucial for practitioners as they showcase a coherent manner to aid fund managers and investors in their investment practices.
For the past couple of years, Machine learning and trading helped by artificial intelligence has drawn growing interest. Here, the approach is used to test the hypothesis that the inefficiency of cryptocurrency industry can be exploited in order to produce anomalous revenue. For the duration between Nov. 2015 and Apr. 2018, daily data for 1, 681 crypto currencies were analyzed. Simple trade techniques supported by state-of -the-art machine learning algorithms are seen to outperform the traditional benchmarks. The results obtained imply that non-trivial, but fundamentally simple, algorithmic processes will help to predict the short-term future of the cryptocurrency market. The popularity of cryptocurrencies had skyrocketed in 2017 due to several consecutive months of super-exponential growth of market capitalization. There are over 1,500 currently recorded cryptocurrencies actively trading today with the cryptocurrencies sitting on more than $300 billion [2], and a total market capitalization of over $800 billion in January 2018. According to a recent survey, between 2.9 and 5.8 million privates as well as institutional investors are in the numerous investment networks and access to markets has become easier over time. In a number of online markets, major crypto currencies can be purchased using fiat currency, and then used in order to purchase less known crypto currencies. The average trading amount is globally exceeding $15bn. About 170 money market funds had been invested in cryptocurrencies since 2017, and Bitcoin futures are launched in order to satisfy the Bitcoin trading and hedging demand for the market. The main objective of the work is to predict the Bitcoin prices, one of the most popular and widely used cryptocurrency which is a source of attraction for many investors as a source of profit or investment. But the market for the cryptocurrencies been volatile since the day it was first introduced. So, the approach towards the survey is to use LSTM RNN and use the available dataset and train the model to give the highest possible accuracy and to provide a real-time price of the Bitcoin for the following days.
Financial Technology (FinTech) has transformed Capital Markets, Payments, Lending and Risk Management through faster processing of financial decisions in real-time and expanding access to financial services digitally. Beneath these user-friendly applications, FinTech platforms rely on foundational Data Structures and Algorithms to provide consistent Latency, Scalable Throughput, Auditing capabilities, and Resilience under adversarial conditions. The purpose of this paper is to review which core Data Structures (Arrays, Linked Lists, Hash Tables, Balanced Trees, Heaps and Graphs) are used to support key FinTech Workload applications (Algorithmic Trading, Fraud Detection, Credit Risk Assessment and Blockchain-based Recordkeeping). In addition, this paper will review the Algorithmic Foundations used to enable common tasks across all these workload applications including Sorting/Searching, Optimization, Statistical Learning and Cryptography, and how Asymptotic Complexity must be evaluated with respect to practical system constraints including Caching Behavior, Concurrency and Failure Modes. There is evidence from the academic literature that Algorithmic Trading can increase liquidity in certain Market Structures while also introducing Systemic Fragility during Stress [1],[2],[3]. And, similarly, there is evidence that Fraud Detection is an inherently adversarial domain where Models and Features must evolve as Attacker Behavior evolves [4],[5]. Lastly, the Paper will discuss several open challenges associated with Scale, Security, Model Governance and Privacy; and evaluate Future-Facing Directions such as Privacy-Preserving Analytics (Federated Learning and Zero-Knowledge Proofs) and Cryptographic Agility to prepare for post-Quantum risk.
Bitcoin en popüler ve yaygın olarak kullanılan dijital para birimidir. Bu nedenle, Bitcoin fiyat hareketinin tahmini finansal piyasalar için büyük önem taşımaktadır. Bitcoin fiyat tahmininde ekonometrik modellerin yanında veri madenciliği yöntemlerinden de faydalanılmaktadır. Veri madenciliğinde kullanılan araç ve yöntemler yardımıyla veriler modellenerek yararlanılacak bilgilere dönüştürülürler. K-Star algoritması veri madenciliği, obje tanımlama ve kontrol sistemleri gibi birçok alanda kullanılmakta olan örnek tabanlı bir yaklaşımdır. Bu çalışmada Makroekonomik değişkenlerin Bitcoin fiyatlarını etkileme seviyeleri, Makine Öğrenme yöntemlerinden Lazy Learning Öğrenmeye Dayalı K-Star Algoritması kullanılarak analiz edilmiştir. Çalışmanın veri seti, bağımlı ve bağımsız değişkenlerin 3 Ocak 2017 - 30 Ocak 2019 yılları arasındaki iş günü bazında 510 adet gözlem değerini içermektedir. Bu gözlemlerin 474 adedi (%93’ü) algoritmanın modellenmesi (eğitim) için, 36 adedi (%7’si) ise sınıflandırma (test) için kullanılmıştır. Modelin Bitcoin fiyatlarını gelecek dönem “yükseliş” mi yoksa “düşüş” mü göstereceğine ilişkin sınıflandırma başarısının %61,1 oranında olduğu, Bitcoin fiyatlarının “yükseliş” göstereceğine ilişkin doğru sınıflandırma başarısının %71,42, “düşüş” göstereceğine ilişkin doğru sınıflandırma başarısının ise %46,66 olduğu tespit edilmiştir. Sonuç olarak Makine Öğrenme Tekniğinin belli bir performans gösterdiği ancak Bitcoin fiyatlarının öngörülebilirliğinin henüz beklentinin altında olduğu ortaya çıkmıştır.
M. Akhil Sai, K. Sarath Chandra Sai, M. Manu Koushik, K. Gowri Raghavendra Narayan
ML and AI-helped exchanging have pulled in developing enthusiasm for as far back as not many years.We examine day-by-day information for different digital currencies over some stretch of time. We show that straightforward exchanging methodologies helped by innovative AI calculations outflank standard benchmarks. We have picked two Machine Learning Algorithms to play out a Comparative Study to foresee cost of a Bitcoin; we have utilized Decision tree regressor and LSTM Algorithms and watched execution of every calculation as far as anticipating the cost of Bitcoin. We saw that Decision tree regressor gives progressively effective and precise outcomes when contrasted with others.
Son yıllarda, bloglar, tweet’ler, forumlar, e-postalar gibi Web 2.0 hizmetleri iletişim kanalı olarak yaygın bir şekilde kullanılmaktadır. Ayrıca sosyal medya; gerek bilgi paylaşımı gerekse istek, şikayet ve dilekler gibi görüşleri belirtmenin en kolay ve en güncel yolu olarak düşünülmektedir. Sosyal medyanın, birçok alana olduğu gibi Bitcoin fiyatlarına olan etkisi de son yıllarda tartışılmaktadır. Bitcoin yıllardır üzerinde durulan ve popülerliği her geçen gün artan bir yatırım aracıdır. Merkezi olmayan bir elektronik para birimi sistemi olan Bitcoin, çok sayıda kullanıcının ilgisini çeken, finansal sistemlerdeki köklü bir değişikliği ifade etmektedir. Bu çalışmada sosyal medyanın, özellikle Twitter kanalından elde edilen tweet’ler bazında, Bitcoin fiyatı ile etkileşimi ortaya konulmuştur. Bunun için 06.10.2018-19.05.2019 tarihleri arasında Twitter kullanıcıları tarafından atılan toplam 2.819.784 tweet üzerinden makine öğrenmesi yöntemlerinden sınıflandırma algoritmaları kullanılarak çeşitli analizler gerçekleştirilmiştir. Bulgular değerlendirildiğinde metin sınıflandırmada %90 ile en yüksek doğruluk oranına sahip olan Yapay Sinir Ağları kullanılmıştır. Ayrıca Bitcoin fiyatları ve sınıflandırılmış olumlu/olumsuz tweet oranları ile ikili korelasyon yapılmıştır. Elde edilen 0,681 korelasyon katsayısı ile pozitif yönde orta üstü kuvvetli ilişki tespit edilmiştir.