A Novel Method for Prediction of Cryptocurrency Price

Abstract

The safe hash method (SHA) 256 and message digest (MD) 5 are used in peer-to-peer transaction arrangements, also known as sophisticated forms of currency, to safeguard data transfers. Prices for Bitcoin are extremely volatile, act erratically, and have reached eccentricity. They are frequently used for initiative and have mostly replaced traditional trading vehicles like metals, bequests, and the stock market. They must be created due to the significance of reliable deciding models in business. However, it is difficult to predict bitcoin\'s price because it is based on other digital currencies. Bitcoin prices have been evaluated by a variety of researchers using machine learning (ML) and deep learning models, in addition to other tendency-based market processes. Changing the price of one type of encrypted money may influence other encrypted types of money because all digital currencies are in the same category. The researchers combined sentiments from Twitter and other online amusement sites to enhance the effectiveness of the framework. DL-Gues, a robust and solid structure for forecasting computerized cash costs that considers its reliance on other cryptography-based currencies and market sentiment, is inspired by this work. Twitter as well as cost reports from Run, Litecoin, and Bitcoin were used in our investigation of the Run cost premise. To determine whether DL-Gather could be applied to more sophisticated monetary standards, we evaluated the ends for the premise for the cost of Bitcoin-Cash using the price data and tweets of Bitcoin, Litecoin, and Bitcoin.

Introduction

I. INTRODUCTION

A type of computer-generated money intended for use in conventional transactions is called a digital currency. It utilizes encryption strategies like SHA-256 and MD-5 to keep banking cooperations hidden. Currently, financial operations can only be carried out with the assistance of third-party organizations like banks; Bitcoin, on the other hand, removes this requirement. Coins are becoming more common in society. It was first introduced in 2008 under the name Bitcoin, with the intention of substituting a universal computerized cash framework for the entire currency trade system [1]. In order to make the framework simple, secure, and decentralized, this newly constructed monetary framework does not include integrated monetary institutions like banks, legislators, or other organizations.

In order to guarantee the system's uniformity and dependability, understanding methods like proof-of-work (PoW) and proof-of-stack (PoS) were developed. Coinage exchange prices were extremely low when it first started. However, its market continues to expand over time due to its instability. As of April 2021, there were approximately 4200 coins on the market, with a market value of $2.23 billion.

With 78% and 12% of the total, respectively, Bitcoin and Ethereum are the most significant donors [2]. Many individuals, businesses, and speculators have made direct or secondary expenditures as a result of the rise in the bitcoin market [3]. The cryptocurrency market is unsettling due to its volatility. The value of coins changes emphatically over the long haul. In less than a decade, Bitcoin's value has increased from $0.08 in 2010 to $64000 in April 2021 [2]. The price of Ethereum increased from $0.67 in January 2018 to $2346 in April 2021 in a scenario comparable to this one [2]. These instances contribute to the explanation of the cryptocurrency market's volatility. Coin prices fluctuate as a result of a number of factors, including traffic, processing complexity, rival digital currency prices, and fame.

The efficient market hypothesis (EMH) and the alternative market hypothesis (AMH) have been utilized by researchers from all over the world to investigate the models and eccentricities of the bitcoin market. The EMH hypothesis states that the prices at which encrypted forms of money are traded are always fair and accurate. Additionally, the associated currency's value will rise as the mining task becomes more difficult. However, in order to address this theory's shortcomings, a new theory known as AMH, which incorporates behavioural finance, was developed. Even if the authors of  are wrong, we might still be able to use EMH to get good results.

II. LITERATURE REVIEW

The use of digital currency has significantly increased as a result of recent advancements in blockchain technology. Digital forms of currency, on the other hand, are not considered a viable venture because of the market's high volatility and cost variation. Due to their fixed nature, most of the digital currency cost-anticipation methods described may not be suitable for ongoing cost anticipation. We present an irregular brain network model to gauge bitcoin costs considering the recently expressed issues. The proposed approach depends on the unpredictable walk hypothesis, which is often utilized in monetary business areas to show stock costs. The suggested approach adds layer-wise sampling to the observed neural network highlight actuations to imitate market volatility. There is also a method for comprehending the example of market reaction in the anticipation model. On Bitcoin, Ethereum, and Litecoin, we demonstrated the Multi-Layer Perceptron (MLP) and Long Short-Term Memory (LSTM) models. The discoveries exhibit that the proposed model outflanks deterministic models [3].

We demonstrate that the five main coins' market efficacy shifts significantly over time. Many bitcoin platforms were inactive prior to 2017. Recent field results support this. However, between 2017 and 2019, the effectiveness of bitcoin exchanges increased. This is refuted by other, more recent findings on the subject. We use a larger sample size than previous studies for one reason. We use a strict efficiency metric to determine whether the efficiency is noteworthy, which is another important reason. Litecoin is the best advanced cash in general, while Wave is the most un-productive [4].

Previously, a document signed by Satoshi Nakamoto under the pen name Nakamoto was how the world learned about Bitcoin. As a result of its huge achievement, a significant number of different coins were created in the years that followed. This dramatic increment is to a great extent because of the markets over the top shakiness, which has started the premium and inclusion of various people, principally for benefit. Bitcoin supporters frequently exchange and learn about news and ideas via web-based entertainment platforms, the most well-known of which is Twitter [5]. We examine how Twitter opinion studies can be used to predict changes in bitcoin prices in this study. We started by gathering price information and tweets from seven of the most popular digital currencies on social media. We then used this information to conduct a felt study using Valence Mindful Word reference for Opinion Thinking (VADER). The Granger Causality test was used first, and then the Augmented Dicky Fuller (ADF) and Kwiatkowski Phillips Schmidt Shin (KPSS) tests were used to evaluate time-series stationarity. Despite the fact that bias in Bitcoin, Cardano, XRP, and Doge appears to be influenced by cost disparities, Ethereum and Polkadot were seen as clearly bullish. Last but not least, Vector Autoregression (VAR) is used to examine the likelihood of price returns, and the predictions for two of the seven coins are astonishingly accurate. Price projections for Polkadot and Ethereum were 99.17% and 99.67%, respectively [6].

It is becoming increasingly common to invest in bitcoin transactions. The forex and equity markets have been compared to the cryptocurrency market. Notwithstanding, because of the capriciousness of bitcoin managing, a conjecture instrument is expected to help purchasers in going with monetary decisions [7]. In today's stock and exchange market predictions, methods based on the calculation of Artificial Neural Networks (ANNs) are frequently employed. There have been a lot of ANN prediction studies on case studies involving currencies and stocks, but none on bitcoin [8]. This study therefore examined a variety of ANN strategies for predicting Bitcoin's market value, one of the most well-known coins. A model that can predict the final value of Bitcoin the following day (following day number) will be created using ANN techniques. Backpropagation neural network (BPNN), genetic algorithm neural network (GANN), genetic algorithm backpropagation neural network (GABPNN), and neuro-evolution of enhancing topologies (NEAT) are the four ANN methods examined in this report. The precision and flightiness of the techniques are looked at. With a MAPE of 1.998 0.038% and a planning season of 347 63 seconds, the assessment uncovered that BPNN is the ideal technique [9].

Bitcoin is now the most widely used sophisticated currency. However, Bitcoin prices have changed a lot, making it hard to estimate. Therefore, the objective of this study is to employ a variety of machine learning techniques to identify the most reliable and practical model for forecasting Bitcoin prices. Using 1-minute stretch transaction data from the Bitcoin trading site bit stamp between January 1, 2012 and January 8, 2018, various backslide models were tested. The Mean Squared Error (MSE) was as low as 0.00002, and the R-Square (R 2) was most likely as high as 99.2% [10].

III. METHODOLOGY

The models and eccentricities of the bitcoin market have been examined by experts from all over the world using the efficient market hypothesis (EMH) and the alternative market hypothesis (AMH). The EMH hypothesis asserts that the prices at which digital currencies are traded are always fair and accurate. Additionally, the associated currency's price will rise due to the extraction project's complexity. However, in order to address this theory's shortcomings, a new theory known as AMH, which incorporates behavioural finance, was developed.

A. Disadvantages

1. The difficulty of the processing job will correspond to an increase in the associated coin's price.
2. Although the authors' method of using EMH is still feasible, it is incorrect.

B. Benefits

1. We investigated and considered showing DL-Gather for two explicit advanced monetary standards to exhibit its power.
2. In terms of predicting bitcoin values, the proposed DL-GuesS algorithm performs better than previous ones. DL-GuesS.  

• Examining the data: We will enter data into the system with this section.
• Information will be given as an expansion to this segment for taking care of.
• Separation of the data into test and training models: Train and test models will be separated by this tool from the data.
• Support is provided for deciding models such as LSTM, GRU, and ARIMA. The estimate isn't set in stone.
• Registration and enrollment of clients: Access to this feature requires enrollment and authentication.
• The intended output will be produced by this program.
• Prediction: The presented actual predicted number.  

• LSTM + CNN: LSTM layers anticipate patterns while CNN layers extract characteristics from incoming data in a CNN-LSTM model. A brief information arrangement known as a period series is typically utilized for sequential data. The DNN method that was chosen was LSTM since it is effective with patterns. When looking for information about a place, like in a photograph, CNN is frequently helpful.
• LSTM (long-term short-term memory): A deep learning architecture incorporates sustain learning, a long-term short-term memory (LSTM), and a false recurrent neural network (RNN). For applications that require time sequences and clusters, LSTMs are a viable option.
• GRU: Kyunghyun Cho et al. in 2014 developed the time-consuming gated recurrent units (GRUs) method for gating cerebrum networks. Although the GRU lacks a result entrance, it performs similarly to a long short-term memory (LSTM) with an ignore doorway.
• Logistic Regression: In machine learning, logistic regression is a categorization method that uses provided dependent factors to predict the likelihood of groups. The logistic regression model then determines the logistic outcome by combining the characteristics of the inputs. with a bias component most of the time).
• Random Forest: Classification and regression problems are frequently solved using the Random Forest method, a guided machine learning technique. We are aware that a forest has a lot of trees, and that the more trees there are, the bigger the forest is.
• DT: For classification and recurrence, a decision tree is a type of non-parametric directed learning method. It has inward hubs, foliage hubs, a base hub, and a sequential tree form.
• SVM: SVM is a backslide-compatible oversaw and a collection of ML algorithms. They become more organized when we label them backslide issues. The goal of the SVM method is to locate a hyperplane that accurately depicts the data centers in an N-layered space.
• MLP: One more progression for ANN with different layers is the multi-facet perceptron. (MLP). A single perceptron is capable of handling straightforward direct tasks, but it is unsuitable for non-straight applications. MLP could be used to deal with these challenging issues.
• Voting Classifier: A voting classifier is an ML grader that makes predictions based on the results of multiple basic models or assessors. Gathering models might be matched prevalence-based decisions for every evaluator return.
• ARIMA: ARIMA models are usually alluded to as ARIMA (p, d, q), where p alludes to the autoregressive model solicitation, d alludes to the level of differencing, and q alludes to the moving-typical model solicitation. Using differencing, ARIMA models turn a non-stationary time series into a stationary one that can be used to predict future characteristics.

IV. IMPLEMENTATION ANALYSIS

ML and deep learning models, in addition to other assessment-based market methods, have been utilized by a variety of specialists to evaluate bitcoin prices. Adjusting the cost of one digital currency may have an impact on other digital currencies because they are all in the same category. The researchers combined sentiments from Twitter and other online amusement sites to improve the accuracy of the framework. DL-Gues, a cross-variable and extreme framework for forecasting computerized cash prices that takes into account its dependence on other encrypted currencies and market sentiments, is inspired by this article.

As an extension to the existing system, we have added few more algorithms by using them which we got Linear Regression as 98.8% with 0.12 MSE Values

IV. FUTURE ENHANCEMENTS

In this proposed model we have included few other algorithms to enhance the accuracy of the prediction, we can also hope that there will be new models in the future that will have more accurate predictions as machine learning keeps on developing by each year which opens new possibilities for the coming generation developers. In the coming years the model may be developed to an extent where it will not only predict the Cryptocurrency prices but will also predict the stock prices by using multiple trained data sets which will help the investors by guiding them to invest their money into stocks without any occurrence of loss.

Conclusion

We looked at the current methods for predicting bitcoin values in this article. Many of them are used by fintech companies to get the most out of bitcoin cost estimations. However, anticipation is difficult due to the market\'s eccentricity and various ward sections. We recommend DL-Surmise, a bitcoin price forecasting half-and-half model that takes price history and current Twitter sentiment into account. In order to comprehend DL-GuesS\' adaptability to two distinct coins, we examined the results, or disaster capacities, in conjunction with previous evaluation. The proposed DL-GuesS method is superior to existing systems when it comes to predicting bitcoin values. DL-GuesS.

References

[1] S. Nakamoto. (2009). Bitcoin: A Peer-to-Peer Electronic Cash System. Cryptography Mailing List. [Online]. Available: https://metzdowd.com [2] CoinMarketCap. (2021). Today’s Cryptocurrency Prices by Market Cap. Accessed: 2021. [Online]. Available: https://coinmarketcap.com/ [3] P. Jay, V. Kalariya, P. Parmar, S. Tanwar, N. Kumar, and M. Alazab, ‘‘Stochastic neural networks for cryptocurrency price prediction,’’ IEEE Access, vol. 8, pp. 82804–82818, 2020. [4] R. Sharma. (2019). Do Bitcoin Mining Energy Costs Influence its Price? Accessed: 2019. [Online]. Available: https://www.investopedia. com/news/do-bitcoin-mining-energy-costs-influence-its-price/ [5] V. L. Tran and T. Leirvik, ‘‘Efficiency in the markets of crypto-currencies,’’ Finance Res. Lett., vol. 35, Jul. 2020, Art. no. 101382. [6] C. Lamon, E. Nielsen, and E. Redondo, ‘‘Cryptocurrency price prediction using news and social media sentiment,’’ SMU Data Sci. Rev., vol. 1, no. 3, pp. 1–22, 2017. [7] A. Radityo, Q. Munajat, and I. Budi, ‘‘Prediction of Bitcoin exchange rate to American dollar using artificial neural network methods,’’ in Proc. Int. Conf. Adv. Comput. Sci. Inf. Syst. (ICACSIS), 2017, pp. 433–438, doi: 10.1109/ICACSIS.2017.8355070. [8] T. Phaladisailoed and T. Numnonda, ‘‘Machine learning models comparison for bitcoin price prediction,’’ in Proc. 10th Int. Conf. Inf. Technol. Electr. Eng. (ICITEE), Jul. 2018, pp. 506–511. [9] M. Wimalagunaratne and G. Poravi, ‘‘A predictive model for the global cryptocurrency market: A holistic approach to predicting cryptocurrency prices,’’ in Proc. 8th Int. Conf. Intell. Syst., Modelling Simulation (ISMS), May 2018, pp. 78–83. [10] I. A. Hashish, F. Forni, G. Andreotti, T. Facchinetti, and S. Darjani, ‘‘A hybrid model for bitcoin prices prediction using hidden Markov models and optimized LSTM networks,’’ in Proc. 24th IEEE Int. Conf. Emerg. Technol. Factory Autom. (ETFA), Sep. 2019, pp. 721–728.

Copyright

Copyright © 2023 M V Nikhil, T. Karthik, Sriram , T. Shirisha, Sunil Bhutada. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.