Summary For the past few years, Bitcoin plays a vital role in both the economical and financial industries. In order to gain a huge return on investment, the investors are eager to forecast the future value of Bitcoin. However, Bitcoin price variation is quite nonlinear and chaotic in nature, so it creates more difficulty in forecasting future value. Researchers found that the multiplicative long short term memory (LSTM) model will be more efficient for predicting those complex variations. So, target mission is about to develop an optimized multiplicative LSTM with an Attention mechanism using Technical Indicators derived from historical data. A modified cuckoo search optimization model is proposed to tune the hyperparameter of the Deep Learning model. This novel optimization algorithm eliminates the local optimum and slower convergence problem of the cuckoo search optimization algorithm. Deibold Mariano test is performed to statistically evaluate the proposed model and it is inferred that the recommended methodology is statistically fit. Regression metrics such as root mean square error, mean square error and mean absolute error has been used for comparative evaluation with related benchmark techniques such as genetic algorithm optimized LSTM (GA–LSTM), particle swarm optimized LSTM (PSO–LSTM) and cuckoo search optimized LSTM (CSO–LSTM). The empirical result shows that the recommended methodology outperforms the taken benchmark models and provides better accuracy.
P. Umamaheswari, S. Abiramasundari, M. Kamaladevi, P. Dinesh
Bitcoin is a type of digital currency or computerized money that is utilised for speculation around the world. Bitcoins are files that are saved in a digital wallet programme on a mobile phone or a PC. Every transaction and its timestamp data are recorded in a common list known as blockchain. In this research, the cost of bitcoin is estimated utilising data mining techniques and machine learning algorithms. The dataset is preprocessed with the use of data mining algorithms, which reduces data noise. Bitcoin's price fluctuates, and it is estimated using long short-term memory (LSTM), a type of neural networking, to extract acceptable patterns for modelling and prediction. Discovering recurring patterns in the bitcoin market is a necessary endeavour in order to achieve optimal bitcoin price functionality. The dataset consists of numerous regularly reported bitcoin price features every year. Linear regression (LR) technique is used to estimate the future cost of bitcoin. Daily price shift with the best possible precision by using the available data is also estimated.
This study analysis forecasting the bitcoin exchange rate against the USD. The dataset selected for this study starts from January 2015 to June 2022. This study's methodology uses autoregressive integrated moving average forecasting (ARIMA). The overall outcomes of this study were gathered from the statistical software Minitab 21.1. The Box Jenkins approaches are also used to predict the best model. To determine the ARIMA model parameter, this study did autocorrelation function (ACF) and partial autocorrelation function (PACF) analyses. According to the Box-Cox transformation method, log transformation was selected. The outcome demonstrates that the seasonal with the regular difference in the Bitcoin exchange rate against the USD is a stationary data series. The forecasting model used in this study is ARIMA (1,1,0) (2,1,1)12. This predicted model is identified through the Mean squared error by comparing the other guessing ARIMA models. After the prediction, 5 Month bitcoin exchange rate against the USD. Investors will be able to estimate the bitcoin exchange rate against the USD with the use of this information, but volatility must also be properly watched. This will aid investors in making better investment decisions and increase profits. In future studies, better consider another exchange rate of BTC and software experts will develop such type of software based on ARIMA models for prediction.
Cryptocurrencies, which occupy a risky position among investment instruments, continue their technological developments day by day with the speed of money transfers and the confidence in the decentralization of production. In this paper, we propose a denoised Autoregressive Integrated Moving Average (ARIMA) and Long Short-Term Memory (LSTM) method for predicting cryptocurrency prices. The daily cryptocurrency price data of Bitcoin (BTC) is collected from a freely available website (cryptocompare.com). For the prediction of BTC/USD price, we considered the average values of daily opening, high, low and closing price as OHLC value. The data set is denoised from white noise using the discrete wavelet transform method (DWT) by VisuShrink thresholding on Daubechies (db4) wavelet at level=5. OHLC and denoised OHLC (DOHLC) values are predicted using ARIMA(p,d,q) and LSTM methods. LSTM hyperparameters are evaluated using 28 different combinations. A pair of Adam-linear optimization and activation function is the best hyperparameter with the lowest mean loss value of 1.42e-03. Finally, DLSTM was found to be the best prediction method according to the Root Mean Squared Error (RMSE) prediction metric: 556.85.
The volatility of cryptocurrencies and exclusivity of crypto communities has made cryptocurrency investment inaccessible for common people. With machine learning, harnessing social media trends that affect price in a random field like cryptocurrency will provide everybody the ability to earn money. Although existing research utilizes sentiment analysis to label posts based solely on English, this project will use NLP to perform stance detection with respect to a certain entity to make predictions. The second part of this project will apply this stance detection to real-world prices, using an RNN to turn stance data into price data. The stance detection model, RoBERTa, reached an accuracy of 80%. An independent price prediction model using an RNN achieved a mean absolute error of $1144, a relatively minimal error considering that the price of crypto reaches $60000. This endeavor proves the difficulty in proving cryptocurrency prices, but the model's steady improvement indicates that future work on social media trends may be promising after all.
Still in the inception of Web 3.0, the present study stands as an important incursion into the future of the Internet and displays how different technologies, Social Media and Cryptocurrencies, can influence each other and the world around them in these new, exciting, times. This paper seeks to find if there is a relationship between Twitter Sentiment and Bitcoin price. To this end, data is collected and preprocessed to construct a dataset consisting of two time series, one for Twitter Sentiment regarding Bitcoin and one for Bitcoin Price. The relationship between the two is studied through a Vector Autoregressive model. The results show that each exerts influence on the other. Twitter Sentiment is quicker to absorb the information, while Bitcoin Price takes longer to intake the most recent events. Future avenues of study are identified, with topic analysis of Bitcoin discourse being highlighted due to its potential to unlock new knowledge.
Andreas Nugroho Sihananto, Anggraini Puspita Sari, Muhammad Eko Prasetyo, Mochammad Yanuar Fitroni · 6 authors
Reinforcement learning as machine learning algorithms can construct software agents, and machines work automatically to determine the superior manners to maximize the algorithm. On the other hand, in recent years, cryptocurrencies are increasingly known because numerous people have used them for investment and trading. People have created automated cryptocurrency trading systems to save time in trading activities. Therefore, researchers are interested in inventing a computerized trading system by implementing a reinforcement learning algorithm. This implementation uses a stable baseline and OpenAi gym with three methods of RNN, such as A2C, ACER, and PPO. The result is that A2C is the best method for low-volume trade like BTC/USDT, while for higher-volume trade in ETH/USDT, the ACER method proved more beneficial. The best BTC/USDT trading method is A2C, with a reward in the testing phase of 0.332. Meanwhile, for ETH/USDT, the best approach is ACER, with the testing phase’s reward is 0.257.
Bitcoin (BTC)-the first cryptocurrency-is a decentralized network used to make private, anonymous, peer-to-peer transactions worldwide, yet there are numerous issues in its pricing due to its arbitrary nature, thus limiting its use due to skepticism among businesses and households. However, there is a vast scope of machine learning approaches to predict future prices precisely. One of the major problems with previous research on BTC price predictions is that they are primarily empirical research lacking sufficient analytical support to back up the claims. Therefore, this study aims to solve the BTC price prediction problem in the context of both macroeconomic and microeconomic theories by applying new machine learning methods. Previous work, however, shows mixed evidence of the superiority of machine learning over statistical analysis and vice versa, so more research is needed. This paper applies comparative approaches, including ordinary least squares (OLS), Ensemble learning, support vector regression (SVR), and multilayer perceptron (MLP), to investigate whether the macroeconomic, microeconomic, technical, and blockchain indicators based on economic theories predict the BTC price or not. The findings point out that some technical indicators are significant short-run BTC price predictors, thus confirming the validity of technical analysis. Moreover, macroeconomic and blockchain indicators are found to be significant long-term predictors, implying that supply, demand, and cost-based pricing theories are the underlying theories of BTC price prediction. Likewise, SVR is found to be superior to other machine learning and traditional models. This research's innovation is looking at BTC price prediction through theoretical aspects. The overall findings show that SVR is superior to other machine learning models and traditional models. This paper has several contributions. It can contribute to international finance to be used as a reference for setting asset pricing and improved investment decision-making. It also contributes to the economics of BTC price prediction by introducing its theoretical background. Moreover, as the authors still doubt whether machine learning can beat the traditional methods in BTC price prediction, this research contributes to machine learning configuration and helping developers use it as a benchmark.
Abstract This paper explores the use of machine learning algorithms and narrative sentiments when applied to the task of forecasting and trading Bitcoin. The forecasting framework starts from the selection among 295 individual prediction models. Three machine learning approaches, namely, neural networks, support vector machines, and gradient boosting approach, are used to further improve the forecasting performance of individual models. By taking data‐snooping bias into account, three different metrics are applied to examine the forecasting ability of each model. Our results suggest that the machine learning techniques always outperform the best individual model whereas the gradient boosting framework has the best performance among all the models. Finally, a time‐varying leverage trading strategy combined with narrative sentiments and volatility is proposed to enhance trading performance. This suggests that the hybrid leverage strategy provides the highest Bitcoin profits consistently among all trading exercises.
Bitcoin has recently been greatly regarded as an investment asset. It is incredibly unpredictable despite being the biggest digital currency. Therefore, accurate forecasting is essential for making investment strategies. This is a difficulty that the latest research effort takes on to construct a revolutionary Bitcoin price prediction model by incorporating new feature engineering and price prediction methods. The original features are first retrieved from the actual Bitcoin data obtained. This work is well-fit and accurate by developing a novel feature computing framework. The proposed decomposed inter-day difference based features and the second order technical indicator are generated within the feature extraction stage. Following that, the developed two-level ensemble classifier is used to accurately forecast the Bitcoin price value using extracted and original features. The two-level ensemble classifier blends the outstanding classifiers support vector machine and artificial neural networks. It is intended to adjust the weight parameter throughout training the ensemble method to accommodate the unpredictability features of Bitcoin prices better. The article presented the novel self-adaptive bat algorithm as a solution. With regard to specific performance metrics, the output of the two-level ensemble classifier is contrasted with that of the current models.
Abstract Cryptocurrencies are distributed digital currencies that have emerged as a consequence of financial technology advancement. In 2017, cryptocurrencies have shown a huge rise in their market capitalization and popularity. They are now employed in today’s financial systems as individual investors, corporate firms, and big institutions are heavily investing in them. However, this industry is less stable than traditional currency markets. It can be affected by several legal, sentimental, and technical factors, so it is highly volatile, dynamic, uncertain, and unpredictable, hence, accurate forecasting is essential. Recently, cryptocurrency price prediction becomes a trending research topic globally. Various machine and deep learning algorithms, e.g., Neural Networks (NN), Gated Recurrent Unit (GRU), Long Short-Term Memory (LSTM), and Bidirectional LSTM (BiLSTM) were utilized to analyze the factors influencing the prices of the cryptocurrencies and accordingly predict them. This paper suggests a five-phase framework for cryptocurrency price prediction based on two state-of-the-art deep learning architectures (i.e., BiLSTM and GRU). The current study uses three public real-time cryptocurrency datasets from “Yahoo Finance”. Bidirectional Long Short-Term Memory and Gated Recurrent Unit deep learning-based algorithms are used to forecast the prices of three popular cryptocurrencies (i.e., Bitcoin, Ethereum, and Cardano). The Grid Search approach is used for the hyperparameters optimization processes. Results indicate that GRU outperformed the BiLSTM algorithm for Bitcoin, Ethereum, and Cardano, respectively. The lowest RMSE for the GRU model was found to be 0.01711, 0.02662, and 0.00852 for Bitcoin, Ethereum, and Cardano, respectively. Experimental results proved the significant performance of the proposed framework that achieves the minimum MSE and RMSE values.
Bhavay Malhotra, Chittaranjan Chandwani, P.N. Agarwala, Suman Mann
A cryptocurrency or ‘coins' is a virtual asset which can be used as an alternative to physical currency via a computer network that is not reliant on the government or any bank, to uphold or maintain it. This has led to the creation of numerous brands of cryptocurrencies. It was only after the boom in their price in 2011 that they began to be regarded as an investment asset. Since these coins are highly volatile in their pricing, there is a need for a good prediction of their closing price on which investment decisions can be made. To address this requirement, this paper studies the relative performances of different machine learning algorithms for a well-known cryptocurrency - the ‘Bitcoin’. The performance measures of different machine learning models were undertaken to get the accuracy of the models for Bitcoin and results were obtained. The results show that the Auto Regressive Integrated Moving Average (ARIMA) is better than the other models and has the least mean absolute error. It is observed that the quality of training data and amount of the dataset used plays an important role for a successful prediction. When comparing the predicted value of Bitcoin through ARIMA with its actual value, the results obtained are found to be comparable for the entire four months of analysis.
In today's world we can see the trend of cryptocurrency is constantly increasing every day. In the financial sector, cryptocurrency has become a huge topic and the right prediction has become more important to gain profits. For determining the right prediction with good accuracy, we performed deep analysis on dataset to understand the market behavior by using different machine learning algorithms like Linear Regression, Random Forest Regressor, Gradient Boosting Regressor, and XGBoost to predict the daily price behavior of top 4 cryptocurrencies like Bitcoin, XRP, Ethereum, and Stellar using these machine learning algorithms. Our experimental result reaches to 95–97 percent validation accuracy.
This paper describes approaches to forecast Ethereum price based on regression analysis which are based on defined in this research list of factors which may affect price. These parameters can be a part of fundamental and technical analysis. In scope of forecasting the nonlinear regression models are used and compared, in couple with prediction of each factor which is used for regression by NeuralProphet. The models’ outputs were retrieved during experiment. Also, experiment includes models tuning to have more accurate result. The data time window for experiment is one year. This paper does not consider influence of political situation and nature cataclysms on cryptocurrency. Also, this research does not include index of openness of countries finance institute. The type of analyzed crypto is decentralized finance. The Java microbenchmark harness is used to calculate time which is spent for models training. Models’ performance is calculated by evaluation of regression metrics: Root mean square error, Mean absolute error, Mean square error, Explained variance.
Cryptocurrencies (CTC) are decentralised digital currency. In the past decade, there has been a massive increase in its usage due to the advancement made in the field of blockchain. Bitcoin (BTC) is the first decentralised CTC which garnered a lot of attention from the media as well as the public due to its ability to sustain the momentum in the market. However, investing in BTC is not the first choice of the investor due to the market’s erratic behaviour, price volatility and lack of a model that could be used to predict its price. In this direction, the present study aims in developing a time-series forecasting model that can efficiently as well as effectively predict the price of Bitcoins. For this purpose three machine learning (ML) models namely Long Short Term Memory (LSTM), Autoregressive Integrated Moving Average method (ARIMA) and Seasonal Autoregressive Integrated Moving Average method (SARIMA) models have been employed which are statistically scrutinised on the basis of the performance metrics namely Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE) and coefficient of determination (R2). The computed value of RMSE, MAPE and $\mathrm{R}^{2}$ for the LSTM model is 1447.648, 3.059% and 0.9702 respectively, ARIMA model is 1288.5, 3.479% and 0.9566 respectively and the SARIMA model is 1802.31, 4.665% and 0.9505 respectively.
In the Present years the requirement for performing secure exchanges online have been designated by the utilization of Virtual tokens called Digital forms of money. Financial backers can bring in cash by performing mining exercises of Digital forms of money like Ethereum and Binance coin or basically selling Ethereum and Binance coin at a benefit rate. Since market costs for digital forms of money depend on market interest, the cost of a specific digital money in contrast with the other Crypto changes generally, As each Crypto coin has its upsides and downsides and as the cost of the coin is differing generally as for the trade paces of coin and the fame of the Crypto coins in public. The work proposes to foresee the future costs of Cryptographic forms of money like Ethereum and Binance Coin utilizing AI draws near. The result of the proposed work is to gauge the costs of Ethereum and Binance Coin and ubiquity in the coming years
Soudeh Javadi Masoudian, Paras M. Kathuria, Nisha S. Gowda, Talha Ali Khan
Cryptocurrency investment has been on trend for the last many years. Many studies have been done on cryptocurrency price prediction based on their parameters and factors. In this research, we concentrate on bitcoin as one of the well-known cryptocurrencies. In this regard, some machine learning approaches predicate the bitcoin price. Neural Networks and Recurrent Neural Networks seem great in bitcoin price prediction and relatively accurate because of their ability to work on time-series data. In this paper, the survey on the performance of LSTM (Long Short-Term Memory), which is one of the Recurrent Neural Networks and is suitable for time-series problems, has proven it can be used for bitcoin price prediction investigation. In this paper, to fulfil this measurement task, a model has been implemented to train and test the dataset; the last year's data was only used because of the fluctuation of the price in recent years. The result shows that LSTM can predict the price remarkably with acceptable accuracy.
Cryptocurrency markets have attracted many interest for global investors because of their novelty, wide on-line availability, increasing capitalization, and potential profits. In the econophysics tradition, we show that many of the most available cryptocurrencies have return statistics that do not follow Gaussian distributions, instead following heavy-tailed distributions. Entropy measures are applied, showing that portfolio diversification is a reasonable practice for decreasing return uncertainty.
Dimitri Mahayana, Shafa Amarsya Madyaratri, Muhammad Fadhl 'Abbas
This research proposes a machine learning-based system that aims to form an investment strategy that is capable of buying and selling cryptocurrencies. Predicting the price movement of crypto assets has a number of challenges, especially due to the high volatility of trading prices. Traditionally this process is carried out by conducting technical analysis using technical indicators which requires extensive knowledge and experience to get maximum profit in cryptocurrency trading. The CRISP-DM methodology was applied to create a classification model to predict the movement of the BTCUSDT cryptocurrency pair using a tree-based classification algorithm with the Gradient Boosting framework, namely the Light Gradient Boosting Machine (LGBM). The Logistic Regression algorithm was used as a comparison. As input data, the klines dataset of the BTCUSDT cryptocurrency pair is used and feature engineering is carried out in the form of labelling (up/down) as well as a number of technical indicators commonly used in trading activities. The prediction results of the classification model are then used in the further evaluation stage by being tested (used for trading) against the original data of BTCUSDT price movements for one month and the Return on Investment (ROI) obtained is evaluated against the Buy and Hold strategy. Based on the evaluation results, the resulting model has a better performance than Logistic Regression, but the model's performance has not been able to exceed the ROI value of the Buy and Hold strategy consistently and still can't generate profit.
Dimitri Mahayana, Elbert Shan, Muhammad Fadhl'Abbas
This research produces a deep reinforcement learning model for algorithmic trading of cryptocurrencies. The model aims to help traders earn greater profits than using traditional strategies. Although traditional strategies can generate profits, they require considerable knowledge, experience, and time to generate optimal profits. Models are trained to trade on the cryptocurrency market. Model inputs are 1 minute interval candlestick data and technical indicators for the BTC/USDT cryptocurrency pair. The model produces an output in the form of a buy, hold, or sell signal. Models are created with the PPO algorithm and a custom environment that follows the gym interface. The performance of the model is compared to the Buy and Hold strategy. The tests that have been carried out show that the model produced in this study still cannot beat the Buy and Hold strategy.