M. Eren Akbiyik, Mert Erkul, Killian Kaempf, Vaiva Vasiliauskaitė · 5 authors
Understanding the variations in trading price (volatility), and its response to exogenous information, is a well-researched topic in finance. In this study, we focus on finding stable and accurate volatility predictors for a relatively new asset class of cryptocurrencies, in particular Bitcoin, using deep learning representations of public social media data obtained from Twitter. For our experiments, we extracted semantic information and user statistics from over 30 million Bitcoin-related tweets, in conjunction with 15-minute frequency price data over a horizon of 144 days. Using this data, we built several deep learning architectures that utilized different combinations of the gathered information. For each model, we conducted ablation studies to assess the influence of different components and feature sets over the prediction accuracy. We found statistical evidences for the hypotheses that: (i) temporal convolutional networks perform significantly better than both classical autoregressive models and other deep learning-based architectures in the literature, and (ii) tweet author meta-information, even detached from the tweet itself, is a better predictor of volatility than the semantic content and tweet volume statistics. We demonstrate how different information sets gathered from social media can be utilized in different architectures and how they affect the prediction results. As an additional contribution, we make our dataset public for future research.
The emergence of Cryptocurrency has long foreshadowed a more accessible exchange market. Cryptocurrency is easy to use and trade, and this ease of access into the market brought newcomers into the Crypto-trading scene. This surge of inexperienced newcomers causes market instability and major loss amongst themselves. AI models, algorithms, and systems have been long used as an important aspect of prediction. However, the use of AI systems is complex. AI tools and systems often use complicated mathematical formulas and are not easily understood. Amongst these AI systems, Fuzzy Rule-Based Systems (FRBSs) has one of the most easily understood displays. With accuracy that rivals of other less-understood methods, such as Neural Network, FRBSs present us a choice that is easily used by users while keeping the interface as basic and simple as possible. This paper aims to study the use of FRBSs using SK-MOEFS (SciKit-Multi Objective Evolutionary Fuzzy System) Python Library in predicting a bull signal or a bear signal in the Cryptocurrency market while still preserving FRBSs user-friendly nature. The fuzzy sets are partitioned as Very Low, Low, Medium, High, and Very High. Then the resulting classification are used to signal whether a Cryptocurrency is bearish or bullish on the current day. The parameter used on the system yields an undesirable result of 53% accuracy with 25 Total Rule Length, however still producing the desired ease-of-use nature of FRBSs.
Bitcoin's dominance in the cryptocurrency market has only increased in recent years. However, it experiences rapid spikes and declines that creates difficulty in predicting its future behavior. Much research has been done to find efficient models that predict with high accuracy, but with limited results. The goal of this study was to determine if higher accuracy can be achieved by focusing on a broader perspective of numeric ranges as opposed to specific time series price predictions. The predictions were concentrated on reporting the expected market direction for the following hour. In using one hour interval trading data and creating discrete classes of levels of hourly changes, five different Machine Learning models were trained and tested. Except for one model, cross validation accuracy ranging from 96-100% was achieved.
Due to the nonlinearity and highly volatile dynamics of the price data of cryptocurrency, classic parametric models show limited success in tracking and prediction. With the rise of deep learning recently, various researches on forecasting the price of cryptocurrency using deep neural network have reported encouraging results in the cases of low volatility. In this study, we propose a hybrid approach which combines the advantages of non-stationary parametric models such as Generalized Autoregressive Conditional Heteroskedasticity (GARCH) with the nonlinear modelling potential of Long-Short Term Memory (LSTM) neural networks. The results show that our hybrid model has a similar predictive performance in terms of MSE, MAE and RMSE but higher metric scores in precision, accuracy and F1 score under optimal hyperparameters. This study reveals that the combination of parametric models like GARCH with deep neural network may come up with better results in cryptocurrency price forecasting especially in the case of highly volatile data or when short data sequences are available. Moreover, the proposed framework can be used also in other applications where high volatility and scarcity of data are the main characteristics.
Aastha Agarwal, S Keerthana, Rahul Reddy, Afraz Moqueem
Bitcoin, Etherium, and Litecoin are among the most extensive market capitalized cryptocurrencies in the present era. With the increased popularity, there is also an increased proclivity of investors towards investing in cryptocurrencies. To gain maximum profits and avoid risks, one needs to analyze the trends and history of the cryptocurrency diligently. This paper put forth various machine learning algorithms to scrutinize cryptocurrencies such as Bitcoin, Etherium, and Litecoin based on multiple trading factors such as open price, close price, volume, market price, history, etc. We have performed various state-of-the-art machine learning to predict the future market value of the cryptocurrencies and derived the performance analysis of the same.
Hasan Kilimci, Mehmet Yigit Yildirim, Zeynep Hilal Kilimci
Bitcoin as a digital cryptocurrency interests the scientists substantially in the areas of computer science, cryptography and economics. In this work, we propose to forecast the last price of Bitcoin Dollar rate in short-term or frequent trading transactions known as day-trading. In addition to statistical indicators such as maximum, minimum, and average prices, technical indicators such as Bollinger band (BB), hour-based moving average (MA), Relative Strength Index (RSI) are also evaluated as a feature set. In order to estimate the price of Bitcoin, different deep and hybrid deep learning methodologies are employed, namely convolutional neural networks (CNNs), long short-term memory networks (LSTMs), convolutional long short-term memory networks (ConvLSTM), CNN Long Short-Term Memory Network (CNN-LSTM). Extensive experiment results exhibit that the usage of ConvLSTM hybrid deep learning model is capable to estimate the price of Bitcoin with 2.4076 of MAPE result.
Cryptocurrency is in great demand today and there is pretty much investment in cryptocurrencies by the investors. There are more than 6000 cryptocurrencies all over the world, which clearly shows that cryptocurrency is a growing investment market. For this reason, investors having ordinary income invest in promising cryptocurrencies with a low market value. However, these investors are often unconsciously investing and making losses. At this point, sensible investments can be made using data analysis methods based on deep learning. Therefore, this study aims to analyze promising cryptocurrencies with deep learning methods. Five promising cryptocurrencies were analyzed with the ensembles of LSTM and single-based LSTM networks. This study revealed that ensembles of LSTM network do not always provide better accuracy performance than the single-based LSTM network in the analysis of promising cryptocurrencies. In other words, these two deep learning methods can be employed to obtain reliable analysis results in promising cryptocurrencies.
Muhammad Shahzad, Laiba Bukhari, Tayyeba Muhammad Khan, S. M. Riazul Islam · 6 authors
Natural Language Processing (NLP) is a challenging and evolving field with the potential of mushroom growth. This technology is expected to assume a pivotal role in bridging the gap between human communication and digital data. In line with that, NLP-driven sentiment analysis has become an attractive research area. On the other hand, bitcoin, arguably the most valuable cryptocurrency, has gained popularity as a major source of investment. In this paper, we propose a framework to perform sentiment analysis on Twitter data. We outline the method and results of predicting bitcoin price for a few days in the future. The framework is expected to be helpful in making informed decisions about our investments and policy based on anticipated future trends.
Since the appearance of bitcoin, the cryptocurrency market has gradually prospered. The price of cryptocurrency often changes rapidly and fluctuates greatly, which makes many investors rush into the market, ignoring the risk behind it. The price prediction of cryptocurrency can not only provide a reference for investors' investment but also reveal its financial laws to a certain extent and improve the risk early warning mechanism as well as improve its stability and security. In this paper, the KNN algorithm is used to forecast the cryptocurrency price, and the KNN model is improved based on the short-term price trend before the trading day. The experimental results show that the improved KNN model has a more accurate prediction result than the logistic regression model and traditional KNN model.
Feature selection reduces the dimension of input variables by eliminating irrelevant features. We propose feature selection techniques based on a genetic algorithm, which is a metaheuristic inspired by a natural selection process. We compare two types of feature selection for predicting a stock market index and cryptocurrency price. The first method is a newly devised genetic filter involving a fitness function designed to increase the relevance between the target and the selected features and decrease the redundancy between the selected features. The second method is a genetic wrapper, whereby we can find the better feature subsets related to KOPSI by exploring the solution space more thoroughly. Both genetic feature selection methods improved the predictive performance of various regression functions. Our best model was applied to predict the KOSPI, cryptocurrency price, and their respective trends after COVID-19.
In this study, we predicted the log returns of the top 10 cryptocurrencies based on market cap, using univariate and multivariate machine learning methods such as recurrent neural networks, deep learning neural networks, Holt’s exponential smoothing, autoregressive integrated moving average, ForecastX, and long short-term memory networks. The multivariate long short-term memory networks performed better than the univariate machine learning methods in terms of the prediction error measures.
Machine learning (ML) algorithms have been widely used to predict future financial trends. It has become a tool for predicting future trends based on what is known beforehand. Like other financial stock markets, cryptocurrency has become a new sensation and challenge for investors to predict its behaviour. However, unlike other financial instruments, cryptocurrency has been renowned because of the difficulty to predict the price due to its volatility behaviour that changes so rapidly and since there is no fundamental economy for its value. This paper presents a performance comparison of two ML algorithms in predicting Ethereum price with non-time series analysis, which are k- Nearest Neighbors (k-NN) and multiple polynomial regression (MPR). The experiment used independent variables from related real-world economic fundamentals such as Dow Jones Index, gold price, oil price, and Ethereum volume. The experiment data was collected from the records from April 2017 until April 2021. For each algorithm, several methods of preprocessing data were used to match all independent data with the dependent data. Three different preprocessing scenarios were also used to find the maximum accuracy model. scenario 1 (feature selection based on correlation matrix), scenario 2 (feature selection based on correlation with the dependent variables and among independent variables), and scenario 3 (scenario 1 extracted with PCA). The performance of the compared methods was evaluated by using MSE and MAE. From the experiment, a comparison of results using two different models with k-NN and multiple polynomial regression is obtained. It is found that k-NN with a hyperparameter K = 2 have the best prediction with MSE = 449.032 and MAE = 14.282 compared with multiple polynomial regression with the best MSE = 13953.96 and MAE = 84.923.
Cryptocurrency is a new sort of asset that has emerged as a result of the advancement of financial technology and it has created a big opportunity for researches. Cryptocurrency price forecasting is difficult due to price volatility and dynamism. Around the world, there are hundreds of cryptocurrencies that are used. This paper proposes three types of recurrent neural network (RNN) algorithms used to predict the prices of three types of cryptocurrencies, namely Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH). The models show excellent predictions depending on the mean absolute percentage error (MAPE). Results obtained from these models show that the gated recurrent unit (GRU) performed better in prediction for all types of cryptocurrency than the long short-term memory (LSTM) and bidirectional LSTM (bi-LSTM) models. Therefore, it can be considered the best algorithm. GRU presents the most accurate prediction for LTC with MAPE percentages of 0.2454%, 0.8267%, and 0.2116% for BTC, ETH, and LTC, respectively. The bi-LSTM algorithm presents the lowest prediction result compared with the other two algorithms as the MAPE percentages are: 5.990%, 6.85%, and 2.332% for BTC, ETH, and LTC, respectively. Overall, the prediction models in this paper represent accurate results close to the actual prices of cryptocurrencies. The importance of having these models is that they can have significant economic ramifications by helping investors and traders to pinpoint cryptocurrency sales and purchasing. As a plan for future work, a recommendation is made to investigate other factors that might affect the prices of cryptocurrency market such as social media, tweets, and trading volume.
More and more people are entering the cryptocurrency market after Bitcoin (BTC) soared to nearly USD 20,000 in 2017. To promote the development of information technology and cryptocurrency marketing, various computerized systems integrating information technology with investment and financing are innovated continuously. In this study, the daily cryptocurrency prices were input to Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM); and the developing trend plots were drawn to predict and analyze the future cryptocurrency prices through deep learning. Finally, the business practices of cryptocurrency investment were modularized based on High-Level Fuzzy Petri Nets (HLFPNs) to make a better investment decision so that all investors can use this decision system to quickly understand the future cryptocurrency trend. The experimental results have shown that this decision system can provide effective investment information to achieve investors’ personal financial goals with the expectation of improving financial situations.
The most significant disturbance now affecting all economies and financial institutions is the digital transformation of economies. The world’s economy and financial institutions are digitizing at an unprecedented rate. Bitcoin is a devolved crypto-currency, or digital asset, that uses blockchain technology to expedite peer-to-peer financial transactions. Price volatility is one of the primary issues with decentralized cryptocurrencies, highlighting the need of examining the underlying price mechanism. Additionally, Bitcoin prices display non-stationary behaviour, meaning that their statistical distribution fluctuates over time. Bitcoin prices are stochastic, and no one set of characteristics can be used to forecast them completely. Nonetheless, academics have demonstrated varying degrees of effectiveness in estimating Bitcoin values using various feature sets. This article explains how to forecast Bitcoin price movements and prices using machine learning approaches. We intend to apply ARIMA, Facebook Prophet and XGBoost techniques for price prediction.
K. Sri Lakshmi Sruthi, D. Ratnagiri, Rudru Jyothika, Salunkhe Sneha · 6 authors
Bitcoin is one of the most popular and valuable cryptocurrencies in the current financial market, attracting traders for investment and thereby opening new research opportunities for researchers. Countless research works have been performed on Bitcoin price prediction with different machine learning prediction algorithms. For the project: relevant features are taken from the dataset having strong correlation with Bitcoin prices and random data chunks are then selected to train and test the model. The random data which has been selected for model training, may cause unfitting outcomes thus reducing the price prediction accuracy. Here, a proper method to train a prediction model is being scrutinised. The proposed methodology is then applied to train a simple Long Short-Term Memory (LSTM) model to predict the bitcoin price for the upcoming 30 days. When the LSTM model is trained with a suitable data chunk, thus identified, sustainable results are found for the prediction. In the end of this project, the work culminates with future improvements. Bitcoin is a kind of Cryptocurrency and now is one of type of investment on the stock market. Stock markets are influenced by many risks of factor. And bitcoin is one kind of cryptocurrency that keep rising in recent few years, and sometimes sudden fall without knowing influence behind it on the stock market. Because it's fluctuations, there's a need and automation tool to predict bitcoin on the stock market. This research study learns how to create model prediction bitcoin stock market prediction using LSTM, LSTM (Long Short-Term Memory) is another type of module provided for RNN later developed and popularized by many researchers, like RNN, the LSTM also consists of modules with recurrent consistency. The Method that we apply on this project, also technique and tools to predict Bitcoin on stock market yahoo finance can predict the result above $ 12600 USD for next days after prediction, in the last section we make conclusions and discuss future works.
David Opeoluwa Oyewola, Emmanuel Gbenga Dada, Juliana Ngozi Ndunagu, Daniel Eneojo Emmanuel
In the wake of recent pandemic of COVID-19, we explore its unprecedented impact on the demand and supply of cryptocurrencies’market using machine learning such as Naïve Bayes (NB), Decision Trees (C5), Decision Trees Bagging (BG), Support Vector Machine (SVM), Random Forest (RF), Multinomial Logistic Regression (MLR), Recurrent Neural Network (RNN), Long Short Term Memory and Noise Bagging (NBG). The study employed Noise filters to enhance the performance of Decision Trees Bagging named NBG. Dataset utilized for this analysis were obtained from the website of Coin Market Cap, including: Binance Coin (BCN), BitCoin Cash (BCH), BitCoin (BTC), BitCoinSV (BSV), Cardano (CDO), Chainlink (CLK), CryptoCoin (CCN), EOS (EOS), Ethereum (ETH), LiteCoin (LTC), Monero (MNO), Stellar (SLR), Tether (TTR), Tezos (TZS), XRP (XRP), and daily data collected from exchange markets platforms spans from 2nd January 2018 to 7th July 2020. Auto encoder was utilized for the labelling of the trading strategies buy-hold-sell.
Irfan Fadil, Muhammad Agreindra Helmiawan, Yanyan Sofiyan
Bitcoin is a cryptocurrency known to have high price fluctuation. Investment depends on price fluctuations which have a high level of risk. Bitcoin investment has these principles. To avoid losses and gain profits, there needs a method that may be used to make forecasts of the price of bitcoin accurately. In this research, Bitcoin price predictions were deployed based on bitcoin price data obtained in the past (Time Series Forecasting) using the method Support Vector Regression. The data retrieved is weekly Bitcoin price data from January 2018 to March 2020. Bitcoin price data is nonlinear, so a kernel is used Radial Basis Function. Meanwhile, the variables of the Support Vector Regression method are optimized using Grid Search Method. The purpose of the research is to determine the accuracy of the Support Vector Regression method by looking at the result of the Mean Absolute Percentage Error value. The research showed that the Mean Absolute Percentage Error value obtained was equal to 10,74 % with parameter value$\mathbf{C=5,} \boldsymbol{\varepsilon=0.004}$, and$\boldsymbol{\gamma=0.07}$. The Mean Absolute Percentage Error value indicates that the prediction results are categorized as a good prediction.
Being the most expensive and most popular cryptocurrency, both the business world and the research community have started to study bitcoin development. However, due to the absence of most government regulation, the price of bitcoin has become uncontrollable, resulting in frequent large fluctuations. Using a dataset from 17 August 2017 to 13 April 2021 the GRU, RNN, and LSTM methods will be compared and implement Grid Search and Random Search to find out which one will do better in this research. Those three methods are considered to be the best method to get a prediction, but it also depends on the model that computer could have. The best result is the GRU with Grid Search method with MAE (Mean Absolute Error) of training 0.0043 and testing about 0.0594.
Bitcoin is an innovative decentralized digital currency without intermediaries. Bitcoin price prediction is a demanding need in the present situation. This paper makes an investigation on the Bitcoin price forecast with a Bi-directional Gated Recurrent Unit (GRU) time series method, combined with opinion mining based on Twitter and Reddit feeds. An hourly basis sentimental analysis through the implementation of Natural Language Processing presents a positive impact of sentimental analysis on the Bitcoin price prediction. For prediction, RNN, long-short memory, GRU has been utilized. Unidirectional and Bi-directional versions of all three networks with and without sentimental analysis were implemented for comparison. Of all the techniques implemented Bi-directional GRU along with sentimental analysis gives a minimum RMSE and Minimum absolute percentage error of 1108.33 and 7.384%. Thus, the framework including Bi-Directional GRU along with Sentimental Analysis provides better results than the State-of-art methods.
Esam Mahdi, Víctor Leiva, Saed Mara’Beh, Carlos Martín-Barreiro
In a real-world situation produced under COVID-19 scenarios, predicting cryptocurrency returns accurately can be challenging. Such a prediction may be helpful to the daily economic and financial market. Unlike forecasting the cryptocurrency returns, we propose a new approach to predict whether the return classification would be in the first, second, third quartile, or any quantile of the gold price the next day. In this paper, we employ the support vector machine (SVM) algorithm for exploring the predictability of financial returns for the six major digital currencies selected from the list of top ten cryptocurrencies based on data collected through sensors. These currencies are Binance Coin, Bitcoin, Cardano, Dogecoin, Ethereum, and Ripple. Our study considers the pre-COVID-19 and ongoing COVID-19 periods. An algorithm that allows updated data analysis, based on the use of a sensor in the database, is also proposed. The results show strong evidence that the SVM is a robust technique for devising profitable trading strategies and can provide accurate results before and during the current pandemic. Our findings may be helpful for different stakeholders in understanding the cryptocurrency dynamics and in making better investment decisions, especially under adverse conditions and during times of uncertain environments such as in the COVID-19 pandemic.
In recent years, there has been an increase in demand for machine learning and AI-assisted trading. To extract abnormal profits from the bitcoin market, the machine learning and artificial intelligence (AI) assisted trading process has been used. Each day, the data gets saved for the specified amount of time. These approaches produce great results when integrated with cutting-edge algorithms. The results of algorithms and architectural structures drive the development of cryptocurrency market. The unprecedented increase in market capitalization has enabled the cryptocurrency to flourish in 2017. Currently, the market accommodates totally 1500 cryptocurrencies, all of which are actively trading. It is always possible to mine the cryptocurrency and use it to pay for online purchases. The proposed research study is more focused on leveraging the accurate forecast of bitcoin prices via the normalization of a particular dataset. With the use of LSTM machine learning, this dataset has been trained to deploy a more accurate forecast of the bitcoin price. Furthermore, this research work has evaluated different machine learning methods and found that the suggested work delivers better results. Based on the resultant findings, the accuracy, recall, precision, and sensitivity of the test has been calculated.