Sentiment Analysis is a technique to determine the tone of a statement using computer software. It is an appliance of a linguistic and computer science intersection that could make a great impact on the business field. Lately, Sentiment Analysis has been used on social media platforms such as Twitter or Facebook to observe the tone of a statement. Examining the tone of tweets using the computer could be time-efficient and precise. SA studies could also be linked to Bitcoin. In this paper, I am using SA results of tweets on a given day to predict changes in the Bitcoin price and its returns. First, I collected the data, which included 31 data points of average sentiment scores and the corresponding 31 Bitcoin prices on the same day. The average sentiment scores were evaluated by VADER from a scale of -1 to 1 (-1 being the statements with the most negative tone, and 1 with the most positive). Then, I used linear regressions to predict Bitcoin price and returns using sentiment scores on the previous day/days. Predicting returns based on sentiments could allow me to find the relationship between Twitter users and Bitcoin and help me better understand the potential challenges. In the end, the predicted price was positively correlated to the sentiment scores the day before. Interestingly, the predicted return was negatively correlated with the sentiment scores and showed less correlation with a M coefficient of -0.86125.
Deep Learning (DL) models are increasingly used for financial forecasting problems, such as price or trend prediction of a financial asset. However, most methods either rely solely on price information or require difficult to implement data harvesting pipelines, e.g., from social media, to deploy them. The main contribution of this paper is a method that exploits sentiment information as a source of additional supervision during the training process, allowing for improving the profitability of the developed strategies compared to baseline agents, while also allowing for operating the agent under partial observability, i.e., without requiring sentiment information as input during inference. As demonstrated in the conducted experiments on the Bitcoin-USD currency pair, this approach can indeed lead to significant improvements in the performance of DL agents, as well as help reduce the overfitting phenomena that often occur when training such agents.
Amadeo Christopher, Kevin Deniswara, Bambang Leo Handoko
This research aims to analyze the calculation of volatility stage from five cryptocurrency products, which are Bitcoin, Ethereum, Binance Coin, Dashcoin, and Litecoin from 1st January 2018 to 1st April 2021 where it consists of calculation of each of the cryptocurrency products' volatility. The research method is a quantitative method by gaining data from Investing.com. Then, analyzing the data using Autoregressive Conditional Heteroscedasticity (ARCH) and Generalized Autoregressive Conditional Heteroscedasticity (GARCH) models. This research aims to know whether ARCH and GARCH models apply to daily life situations in the field. The result shows that the data from ARCH and GARCH models are not suitable on daily basis. Further research should calculate cryptocurrency products to use differentiated GARCH models, such as GJR-GARCH or GARCH-MIDAS. It is also better to calculate the volatility of cryptocurrency products annually. According to some thesis, the volatility cryptocurrency products are more suitable to calculate annually than daily.
Gian Pietro Bellocca, Giuseppe Attanasio, Luca Cagliero, Jacopo Fior
Cryptocurrency trading has become more and more popular among private investors. According to recent studies, the momentum effect influences the underlying market. Quantitative trading systems can leverage momentum indicators to open and close trading positions. However, existing approaches that exploit the momentum effect in cryptocurrency trading do not rely on machine learning. Since these systems are based on human generated rules they are not suited to highly volatile market conditions, which are quite common in cryptocurrency markets. This paper proposes to leverage machine learning approaches to automatically detect the momentum effect in cryptocurrency market data. For each cryptocurrency it estimates the likelihood of being affected by the momentum effect on the next trading day as well as the momentum direction. A backtesting session, performed on three very popular cryptocurrencies, shows that the machine learning models are able to predict, to a good approximation, short-term price volatility thus reducing the number of false trading signals and increasing the return on investments compared to state-of-the-art approaches.
Throughout the history of modern finance, very few financial instruments have been as strikingly volatile as cryptocurrencies. The long-term prospects of cryptocurrencies remain uncertain; however, taking advantage of recent advances in neural networks and volatility, we show that the trading algorithms reinforced by short-term price predictions are bankable. Traditional trading algorithms and indicators are often based on mean reversal strategies that do not advantage price predictions. Furthermore, deterministic models cannot capture market volatility even after incorporating price predictions. Thus motivated by these issues, we integrate randomness in the price prediction models to simulate stochastic behavior. This paper proposes hybrid trading strategies that take advantage of the traditional mean reversal strategies alongside robust price predictions from stochastic neural networks. We trained stochastic neural networks to predict prices based on market data and social sentiment. The backtesting was conducted on three cryptocurrencies: Bitcoin, Ethereum, and Litecoin, for over 600 days from August 2017 to December 2019. We show that the proposed trading algorithms are better when compared to the traditional buy and hold strategy in terms of both stability and returns.
This paper discusses, trying to accurately assess the price of Bitcoin by looking at different parameters affects the value of Bitcoin. In our work, we focus on understanding and seeing the evolution of Bitcoin daily market, a1 and gaining intuition in the most relevant aspects surrounding the Bitcoin price. In the meantime, market capitalization of publicly traded cryptocurrencies exceeds $ 230 billion. The most important cryptocurrency, Bitcoin, is used primarily as a digital value store, and its pricing opportunities have been extensively considered. These features are described in more detail in the following paragraph: details of the main Bitcoin, as described in the paper. Bitcoin is the most expensive digital currency in the market. However, Bitcoin prices have been highly volatile, making it difficult to forecast. As a result, the goal of this research is to find the most efficient and accurate model for predicting Bitcoin prices using various machine learning algorithms. Several regression models with scikit-learn and Keras libraries were tested using 1-minute interval trading data from the Bitcoin exchange website bit stamp from January 1. 2012 to January 8, 2018. The best results showed a Mean Squared Error (MSE) as low as 0.00002 and an R- Square (R2) as high as 99.2 percent.
Stock market price and cryptocurrency price prediction is a very challenging task. We are proposing dynamic algorithms which make use of LSTM and another time Series algorithm, i.e., prophet and we have various trained models on these two algorithms. We will make use of this dynamic algorithm which will self-evaluate different datasets and different pretrained models and will provide us with the best possible output for different test cases. For the longer duration, we are just focusing on up and down, but for the small duration, we are focusing on price-related accuracy. The main and challenging work is to deal with the dynamic dataset, so we require some dynamic algorithm for this.
YUTENG LIU, YUXUAN TIAN, Tianxing Zhou, HONGZHOU WANG
Since the rise of Data Analysis, forecasting of price markets has never stopped and there are numerous forecasting methods, but most of them are only for a single price data.We have chosen bitcoin and gold as the subjects of our study, addressing the multi-objective related prediction problem, explores the volatility relationship between gold and bitcoin to improve its forecasting accuracy, and in doing so, we establishes multiple prediction models,and determines the relationship between prediction accuracy and prediction range.
With the proliferation of pump-and-dump schemes (P&Ds) in the cryptocurrency market, it becomes imperative to detect such fraudulent activities in advance to alert potentially susceptible investors. In this paper, we focus on predicting the pump probability of all coins listed in the target exchange before a scheduled pump time, which we refer to as the target coin prediction task. Firstly, we conduct a comprehensive study of the latest 709 P&D events organized in Telegram from Jan. 2019 to Jan. 2022. Our empirical analysis reveals some interesting patterns of P&Ds, such as that pumped coins exhibit intra-channel homogeneity and inter-channel heterogeneity. Here channel refers a form of group in Telegram that is frequently used to coordinate P&D events. This observation inspires us to develop a novel sequence-based neural network, dubbed SNN, which encodes a channel's P&D event history into a sequence representation via the positional attention mechanism to enhance the prediction accuracy. Positional attention helps to extract useful information and alleviates noise, especially when the sequence length is long. Extensive experiments verify the effectiveness and generalizability of proposed methods. Additionally, we release the code and P&D dataset on GitHub: https://github.com/Bayi-Hu/Pump-and-Dump-Detection-on-Cryptocurrency, and regularly update the dataset.
Ali Raheman, Anton Kolonin, Igors Fridkins, Ikram Ansari · 5 authors
In this paper, we explore the usability of different natural language processing models for the sentiment analysis of social media applied to financial market prediction, using the cryptocurrency domain as a reference. We study how the different sentiment metrics are correlated with the price movements of Bitcoin. For this purpose, we explore different methods to calculate the sentiment metrics from a text finding most of them not very accurate for this prediction task. We find that one of the models outperforms more than 20 other public ones and makes it possible to fine-tune it efficiently given its interpretable nature. Thus we confirm that interpretable artificial intelligence and natural language processing methods might be more valuable practically than non-explainable and non-interpretable ones. In the end, we analyse potential causal connections between the different sentiment metrics and the price movements.
Leonardo Kanashiro Felizardo, Francisco Caio Lima Paiva, Catharine de Vita Graves, Élia Yathie Matsumoto · 7 authors
The interdisciplinary relationship between machine learning and financial markets has long been a theme of great interest among both research communities. Recently, reinforcement learning and deep learning methods gained prominence in the active asset trading task, aiming to achieve outstanding performances compared with classical benchmarks, such as the Buy and Hold strategy. This paper explores both the supervised learning and reinforcement learning approaches applied to active asset trading, drawing attention to the benefits of both approaches. This work extends the comparison between the supervised approach and reinforcement learning by using state-of-the-art strategies with both techniques. We propose adopting the ResNet architecture, one of the best deep learning approaches for time series classification, into the ResNet-LSTM actor (RSLSTM-A). We compare RSLSTM-A against classical and recent reinforcement learning techniques, such as recurrent reinforcement learning, deep Q-network, and advantage actor–critic. We simulated a currency exchange market environment with the price time series of the Bitcoin, Litecoin, Ethereum, Monero, Nxt, and Dash cryptocurrencies to run our tests. We show that our approach achieves better overall performance, confirming that supervised learning can outperform reinforcement learning for trading. We also present a graphic representation of the features extracted from the ResNet neural network to identify which type of characteristics each residual block generates.
Cryptocurrencies can be considered as mathematical money. As the most famous cryptocurrency, the Bitcoin price forecasting model is one of the popular mathematical models in financial technology because of its large price fluctuations and complexity. This paper proposes a novel ensemble deep learning model to predict Bitcoin’s next 30 min prices by using price data, technical indicators and sentiment indexes, which integrates two kinds of neural networks, long short-term memory (LSTM) and gate recurrent unit (GRU), with stacking ensemble technique to improve the accuracy of decision. Because of the real-time updates of comments on social media, this paper uses social media texts instead of news websites as the source data of public opinion. It is processed by linguistic statistical method to form the sentiment indexes. Meanwhile, as a financial market forecasting model, the model selects the technical indicators as input as well. Real data from September 2017 to January 2021 is used to train and evaluate the model. The experimental results show that the near-real time prediction has a better performance, with a mean absolute error (MAE) 88.74% better than the daily prediction. The purpose of this work is to explain our solution and show that the ensemble method has better performance and can better help investors in making the right investment decision than other traditional models.
Anticipating price developments in financial markets is a topic of continued interest in forecasting. Funneled by advancements in deep learning and natural language processing (NLP) together with the availability of vast amounts of textual data in form of news articles, social media postings, etc., an increasing number of studies incorporate text-based predictors in forecasting models. We contribute to this literature by introducing weak learning, a recently proposed NLP approach to address the problem that text data is unlabeled. Without a dependent variable, it is not possible to finetune pretrained NLP models on a custom corpus. We confirm that finetuning using weak labels enhances the predictive value of text-based features and raises forecast accuracy in the context of predicting cryptocurrency returns. More fundamentally, the modeling paradigm we present, weak labeling domain-specific text and finetuning pretrained NLP models, is universally applicable in (financial) forecasting and unlocks new ways to leverage text data.
Mamoona Zahid, Farhat Iqbal, Abdul Raziq, Naveed Sheikh
Using the high-frequency data of Bitcoin, this study aims to model the time-varying volatility identified in the residuals of the heterogeneous autoregressive (HAR) model of realized volatility using the symmetric, asymmetric and long-memory generalized autoregressive conditional heteroscedastic models (GARCH) models. We further extended these models by incorporating jumps and continuous components in the realized volatility estimators and investigating the impact of the inverse leverage effect. The Diebold Mariano and model confidence set test confirm that the forecasting performance of HAR-type models can be effectively improved by these innovations. The long memory HAR-GARCH model with jumps and continuous components provided better forecasting accuracy for Bitcoin volatility as compared to other realized volatility models. The findings of this study may benefit individual investors and risk managers who wish to minimize risks and diversify their portfolios to maximize profits in Bitcoin’s investment.
Cryptocurrency is an excellent investment opportunity. Predicting the prices of cryptocurrencies is a problem that has been around since the origin of cryptocurrencies. Many machine learning approaches are proposed to solve the problem but none of them couldn't solve the problem with expected accuracy. Recent developments in deep learning have proven that using deep learning is one of the best approaches for predicting the prices of cryptocurrencies. In this paper, how LSTM networks are an upgrade over the previous solutions (regression, traditional neural networks, simple recurrent neural networks) and how they can be implemented for predicting the price of any cryptocurrency have been demonstarted.
The objectives of this research are 1) to study the relationship of people interested in investing in cryptocurrencies; 2) to identify people interested in cryptocurrencies. This research study is quantitative research by survey research using an online questionnaire data collection method of 402 respondents. The research results found that Most of them are Generation Z people interested in investing in cryptocurrencies and low to middle-income groups, most of whom are interested in BTC and ETH, the well-known crypto-currency groups. The analysis was then divided by K-Means Clustering analysis into three groups, namely "Not interested in cryptocurrencies", "Moderate-interest" and "Risk-takers" that interested in investing in cryptocurrencies. The last group has a selection to invest in a variety of cryptocurrencies, especially SOL, BNB and ADA, which are smart contract coins.
Cryptocurrencies gaining popularity as a digital currency in today’s world. The market is altering dramatically as a result of the rapid increase in investment in digital currency. The ability to accurately forecast and change the behavior of digital currency is becoming a critical aspect in today’s digital environment. This prompted us to investigate bitcoin trading and investor adoption in the market. We began by examining and analyzing the costs of currencies, and after reviewing numerous methods, one of them proved to be the ideal technique for our data representation, taking into account various criteria such as date, time, and year. A time series forecasting technique has been implemented for forecast the future profit or loss in currencies.
K Dhinakaran, Baby Shamini P, J Divya, C Indhumathi · 5 authors
One of the most valuable currency across the globe right now is Cryptocurrency. Apart from being highly valued, its value increased from approximately 1 dollar in 2010 to 57521,576 in 2021 (for Bitcoin). Again, in recent years, it has attracted considerable attention in a variety of fields, including economics and computer science. The former focuses on studies to determine price fluctuations and its future prices for factors that determine how it will affect the market. The latter mainly focuses on its vulnerabilities, scalability and other techno-cryptocurrency issues. Its aim is to reveal the advantage of the traditional Autoregressive Integrative Moving Average (ARIMA) model in estimating the future value of cryptocurrency by analysing the price time series over a period of 3 years. On one hand, the factual studies show that the conduct of the time series is nearly unchanged, this simple scheme is efficient in sub-periods for the most part when it is used for short-term prediction, the further investigation in Cryptocurrency prediction of the price using an ARIMA model which has been trained over the whole dataset, as well as a limited part of the history of the Cryptocurrency price, with the input of length being w. The interaction of the prediction accuracy and choice of window size is well highlighted in the work.
Predicting the future price of the currency has always been considered one of the most challenging issues. In this paper, we utilize different artificial neural networks (ANNs), including Feedforwardnet, Fitnet, and Cascade networks, and predict the future price of Bitcoin. This paper discusses how a combination of technical attributes, like price-related and lagged features, as inputs of the neural networks, are used to raise the prediction capabilities that directly impact into the final profitability. For empirical analysis, this paper uses the data of the Bitcoin price for a period of 9 months (1.1.2018 - 30.9.2018) available on http://www.coindesk.com. Using a ten-fold cross-validation method, this paper finds the optimal number of hidden neurons for different train functions in each ANN based on error measures, including mean squared error (MSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). Then, the Bitcoin price is predicted, and results are compared based on the amount of R to find out which ANN leads to a better prediction. Finally, this paper concludes that the Fitnet network with trainlm function and 30 hidden neurons outweighs the others. This paper assesses the models’ performance and how specific setups produce principled and stable predictions for beneficial trading.