This paper presents an empirical verification of the effectiveness and usefulness of investment diversification using the main stock exchange indices and Bitcoin. The objective is to determine the effects applying the Markowitz portfolio optimization theory, i.e., the advantages of applying the modern portfolio theory for institutional investors. The research offers an answer to the following question: what are the advantages and disadvantages of using Bitcoin in portfolio optimization? The paper contributes to the representation of the reach and limitations of the modern portfolio theory for institutional investors. The conclusion is that rational behaviour of institutional investors requires consideration of portfolio optimization using the Markowitz model, because it is possible to create portfolios which, on the basis of historical returns, provide desired returns alongside certain risks. The methodology includes the analysis of high frequency data, i.e., daily trading data were used. The results indicate that the use of the Markowitz portfolio selection method, with all its limitations, is desirable, possible and applicable, but that it entails serious flaws in the sense of neglecting transaction costs, foreign exchange differences and the real value in the stock market. The results of the research show that Bitcoin is a good source of diversification in a portfolio that contains traditional financial instruments both for the risk-averse investor as well as for those investors who have a greater appetite for risk. The conclusion is that rational behavior of institutional investors requires consideration of investing in Bitcoin using the Markowitz model. However, given the high degree of volatility, investors should be very careful when making decisions about including Bitcoin in the portfolio.
Ahmet Faruk Aysan, Ali Yavuz Polat, Hasan Tekin, Ahmet Semih Tunalı
This study aims to investigate the effect of fear sentiment with a novel data set on Bitcoinâs return, volatility and transaction volume. We divide the sample into two subperiods in order to capture the changing dynamics during the COVID-19 pandemic. We retrieve the novel fear sentiment data from Thomson Reuters MarketPsych Indices (TRMI). We denote the subperiods as pre- and post-COVID-19 considering January 13th, 2020, when first COVID-19 confirmed case was reported outside China. We employ bivariate vector autoregressive (VAR) models given below with lag-length k, to investigate the dynamics between Bitcoin variables and fear sentiment. Bitcoin market measures have dissimilar dynamics before and after the Coronavirus outbreak. The results reveal that due to the excessive uncertainty led by the outbreak, an increase in fear sentiment negatively affects the Bitcoin returns more persistently and significantly. For the post-COVID-19 period, an increase in fear also results in more fluctuations in transaction volume while its initial and cumulative effects are both negative. Due to extreme uncertainty caused by the COVID-19 pandemic, investors may trade more aggressively in the initial phases of the shock.
Starting from the earlier civilisation till date, money plays a crucial part in the transfer of goods and services. With this digital world, the money also changes its faces from paper money to digital currency called cryptocurrency without any central bank, which runs on top of the technology called blockchain. The trendiest cryptocurrency is bitcoin. Forecasting the daily price is a challenging task due to its nonlinearity. Most of the researchers tried to predict using various statistical and machine learning models which were not satisfactory because of its large dataset with more noise. The intention is to design a deep learning multiplicative long short-term memory model to estimate the price of bitcoin with an attention mechanism using technical indicators which gives better accuracy and a very less error rate. The proposed model is compared with some existing models, say long short-term memory, peephole, gated recurrent unit and multiplicative long short-term memory on the presence and absence of technical indicators. The comparative result shows that the proposed model outperforms the existing models in terms of mean square error, root mean square error and mean absolute error when evaluated with two benchmark datasets.
This work is a comparative study of different univariate and multivariate time series predictive models as applied to Bitcoin, other cryptocurrencies, and other related financial time series data. ARIMA models, long regarded as the gold standard of univariate financial time series prediction due to both its flexibility and simplicity, are used a baseline for prediction. Given the highly correlative nature amongst different cryptocurrencies, this work aims to show the benefit of forecasting with multivariate time series modelsâprimarily focusing on a novel parameter optimization of VARIMA models outlined in this paper. These models are trained on 3 years of historical data, aggregated from different cryptocurrency exchanges by Coinmarketcap.com, which includes: daily average prices and trading volume. Historical time series data of traditional market data, including the stock Nvidia, the de facto leading manufacture of gaming GPUâs, is also analyzed in conjunction with cryptocurrency prices, as gaming GPUâs have played a significant role in solving the profitable SHA256 hashing problems associated with cryptocurrency mining and have seen equivalently correlated investor attention as a result. Models are trained on this historical data using moving window subsets, with window lengths of 100, 200, and 300 days and forecasting 1 day into the future. Validation of this prediction against the actually price from that day are done with following metrics: Directional Forecasting (DF), Mean Absolute Error (MAE), and Mean Squared Error (MSE).
Ahmed Saied El-Berawi, Mohamed Belal, Mahmoud Mahmoud Abd Ellatif
This paper proposes a deep learning based predictive model for forecasting and classifying the price of cryptocurrency and the direction of its movement. These two tasks are challenging to address since cryptocurrencies prices fluctuate with extremely high volatile behavior. However, it has been proven that cryptocurrency trading market doesnât show a perfect market property, i.e., price is not totally a random walk phenomenon. Based upon this, this study proves that the price value forecast and price movement direction classification is both predictable. A recurrent neural networks based predictive model is built to regress and classify prices. With adaptive dynamic features selection and the use of external dependable factors with a potential degree of predictability, the proposed model achieves unprecedented performance in terms of movement classification. A naĂŻve simulation of a trading scenario is developed and it shows a 69% profitability score a cross a six months trading period for bitcoin.
Due to the inherent chaotic and fractal dynamics in the price series of Bitcoin, this paper proposes a twoâstage Bitcoin price prediction model by combining the advantage of variational mode decomposition (VMD) and technical analysis. VMD eliminates the noise signals and stochastic volatility in the price data by decomposing the data into variational mode functions, while technical analysis uses statistical trends obtained from past trading activity and price changes to construct technical indicators. The support vector regression (SVR) accepts input from a hybrid of technical indicators (TI) and reconstructed variational mode functions (rVMF). The model is trained, validated, and tested in a period characterized by unprecedented economic turmoil due to the COVIDâ19 pandemic, allowing the evaluation of the model in the presence of the pandemic. The constructed hybrid model outperforms the single SVR model that uses only TI and rVMF as features. The ability to predict a minute intraday Bitcoin price has a huge propensity to reduce investorsâ exposure to risk and provides better assurances of annualized returns.
Purpose Motivated by the lure of cryptocurrencies for retail investors, whose concentrated holdings are particularly exposed to price crash risk, this paper aims to study the relationship between investor attention and crash risk for a range of cryptocurrencies. Design/methodology/approach This study adopts a quantile regression approach to determine the effect of investor attention on crash risk. Crash risk is measured using the negative coefficient of skewness and down up volatility. Findings This study finds that the connection is concentrated in the tails of the crash risk distribution. Investor attention has a positive relationship with crash risk when crash risk is low (below-median quantiles) and negative when crash risk is high (above-median). The findings are consistent for different measures of crash risk, for alternate internet searches and for a panel of large cryptocurrencies in addition to Bitcoin. This study also notes seasonality in crash risk, with higher crash risk during the JuneâAugust period and lower crash risk in the Halloween period that runs from November to April. Originality/value The results provide insights that are not apparent in previous analyses of cryptocurrency price crash risk. The results are particularly important for retail investors, who constitute a large portion of the cryptocurrency market, as they tend to hold concentrated investments and so a price crash of a single asset may have a large bearing on their wealth.
In this article, we establish a method to detect and formulate price bubbles in the cryptocurrency markets. This method identifies abnormal crashes through violations of the exponential decaying property. Confirmations of bubble bursts within these anomalies are obtained through wavelet analysis. By decomposing the cryptocurrency price into the high-frequency and low-frequency factors, we distinguish the price regimes versus the periods with bubbles and crashes in both time and frequency domains. In addition, we apply the log-periodic power law model to fit the bubble formation. In the analysis of eight cryptocurrenciesâBitcoin, Ethereum, Litecoin, Antshares, Ethereum Classic, Dash, Monero, and OmiseGOâfrom 15 May 2018 to 28 November 2022, we identify 24 bubbles. Some of them exhibit a significant and strong exponential growth pattern.
Fırat Akba, Ä°Ìhsan Tolga Medeni, Mehmet Serdar GĂŒzel, I. N. Askerzade
Today, there are constant changes in terms of securities in stock markets. In these stock market investments, investors use fundamental analysis tools and indicators very widely. In this way, it is possible to have some knowledge of the situations experienced in the markets and to make a profit. In this study, manipulations on Bitcoin are discussed. Popular machine and statistical forecasting methods have been used to detect these manipulations and the road maps to be followed in order to be detected in the most successful way have been shared. Social media sentiments, which were thought to have an effect on manipulations during the studies, were also evaluated with the most advanced text analysis methods and evaluated together with these price changes. The allegations that the prediction methods carried out before the crisis were more successful were investigated. The Covid-19 pandemic was evaluated as a period of global crisis and the studies that might be relevant were examined. It would not be wrong to say that the actors that make big gains in the stock markets are the ones that determine the direction of the stock market. The manipulation periods of the market actors to be successful in the virtual money markets have been tried to be verified by various estimation methods. These estimations can achieve up to F1score of 93% success according to our experimental result. Besides, it is stated that accounts with the highest volume of transactions in the periods, when anomalies were detected, were labeled as potential manipulators.
Jakob Albers, Mihai Cucuringu, Sam Howison, Alexander Y. Shestopaloff
In light of micro-scale inefficiencies induced by the high degree of fragmentation of the Bitcoin trading landscape, we utilize a granular data set comprised of orderbook and trades data from the most liquid Bitcoin markets, in order to understand the price formation process at sub-1 second time scales. To achieve this goal, we construct a set of features that encapsulate relevant microstructural information over short lookback windows. These features are subsequently leveraged first to generate a leader-lagger network that quantifies how markets impact one another, and then to train linear models capable of explaining between 10% and 37% of total variation in $500$ms future returns (depending on which market is the prediction target). The results are then compared with those of various PnL calculations that take trading realities, such as transaction costs, into account. The PnL calculations are based on natural $\textit{taker}$ strategies (meaning they employ market orders) that we associate to each model. Our findings emphasize the role of a market's fee regime in determining its propensity to being a leader or a lagger, as well as the profitability of our taker strategy. Taking our analysis further, we also derive a natural $\textit{maker}$ strategy (i.e., one that uses only passive limit orders), which, due to the difficulties associated with backtesting maker strategies, we test in a real-world live trading experiment, in which we turned over 1.5 million USD in notional volume. Lending additional confidence to our models, and by extension to the features they are based on, the results indicate a significant improvement over a naive benchmark strategy, which we also deploy in a live trading environment with real capital, for the sake of comparison.
V. R. Niveditha, Karthik Sekaran, K. Amandeep Singh, Sandeep Kumar Panda
Internet of things is the concept of establishing relationships and interactions with other connected devices through a network to reach a specific objective. The collected data from devices could be transformed into valuable insights by applying some intelligent learning algorithms. A distributed, permission less ledger called IOTA (MIOTA) manages micro transactions between multiple devices with the help of IoT. It provides the information about the transactions made with the specific type of crypto currency in the market. In this paper, an effective crypto currency price prediction model is proposed to identify the fluctuations in currency value from the past one-year data. Wolf search optimisation algorithm selects the best performing feature subset. Bidirectional long short-term memory (BiLSTM) model is employed to train and validate the data captured from the feature selection process. The proposed model attained 93% accuracy, significantly higher than the existing methods, portraysits significance and efficacy.
Nishant Jagannath, Tudor Barbulescu, Karam M. Sallam, Ibrahim Elgendi · 8 authors
Bitcoin generates a massive amount of data every day due to its innate transparency and capacity of operating completely decentralised. In this paper, we introduce on-chain metrics derived from data on the bitcoin network that enable us to describe the state and usage of the underlying network. Based on their characteristics, we classify them into user, miner, exchange activities and run a correlation analysis with the price to understand the dynamics of bitcoin's price and its underlying mechanics. Using the correlated data, we develop a deep learning model. However, determining the best values of parameters in a deep learning model can be a very challenging and time-consuming task. Hence, we propose a self-adaptive technique using a jSO optimization algorithm to find the best values of these parameters to accurately predict the price of bitcoin. Compared to traditional LSTM model, our approach is highly accurate and optimised with a minimum error rate.