See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/347453695

Predictive Analysis of Cryptocurrency using Machine Learning with Blockchain

technology
Article in Machine Learning · December 2020

CITATIONS READS
0 773

1 author:

Sudheer Palakurla
Bournemouth University
1 PUBLICATION 0 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Research Methods and Professional Issues View project

All content following this page was uploaded by Sudheer Palakurla on 18 December 2020.

The user has requested enhancement of the downloaded file.

 Predictive Analysis of Cryptocurrency using
Machine Learning with Blockchain technology
Sudheer Palakurla
Department of Computing and Informatics
Bournemouth University
Bournemouth, England, United Kingdom.

Abstract: this information is depending upon the

third-party financial institutions which are
Cryptocurrency is one of the fast-growing responsible for the theft or loss of security in
[1] st
digital money in the present world. In this the information systems. Now in this 21
study, Predictive analysis of cryptocurrency century, new digital currencies such as
provides the right methodology for explicit cryptocurrencies are evaluating day by day
predictions of cryptocurrency markets helps leaving behind the actual currency in the
[11]
the investors to make the right investment world.
decisions to attain profits. Using Machine
Learning algorithms such as Bayesian neural Based on the numerous numbers of
networks, supervised learning and random investors have an interest in purchasing of
these cryptocurrencies made a major impact
forests will be able to analyse the price
[1]
fluctuations of the cryptocurrencies through on the economy. In the next decade,
historical data managed by Blockchain cryptocurrency will grow exponentially
technology. This results in the high precision higher in the market. Many cryptocurrencies
rate of the prediction analysis of such as Bitcoin, Ethereum, Litecoin, Stellar,
Ripple and many more are presently growing
cryptocurrency using Machine Learning
[2]
algorithms along with Blockchain technology. rapidly in the market.

For short term investment plans most of the

Keywords: Cryptocurrency, Supervised financial investors who are interested in
learning, Random forest, sentiment analysis, cryptocurrency would invest for their
Machine Learning, Blockchain Technology. benefits. Bitcoin is one of the popularly
known cryptocurrencies which is used for
Introduction: [7]
diversification purposes.

In the past decade, the banks recorded the Cryptocurrencies are developed with
transactions using paper-based which are Blockchain technology which maintains all
now adapting to digital transactions for all the transfers in distributed ledger format
the financial assets such as bank deposits, consists of records also known as blocks.
bonds or equities are digitally recorded. All These blocks will record transactions of the
data, timestamps and previous transactions currencies such as cryptocurrencies are not
involved in the block which cannot be been managed by any government or central
altered later. Blockchain technology along bank depends mostly on the social media
with machine learning is now reliable and its trends, the prediction of prices are
[16]
increase in the present world is phenomenal. calculated.
Blockchain is one of the best technologies to
record the information of data which is In most of the papers, the bitcoin price
prediction in the cryptocurrency market is
impenetrable to manipulate. It made a big
[13] [14] [15]
difference in the invention and trading of calculated. References using the
cryptocurrency. With the use of Blockchain blockchain technology features to predict
technology along with machine learning the bitcoin price with the change of the
prices which results in success in most of the
algorithms made it easy for predictive and
cases. There is a record of 64% accuracy of
descriptive methods to identify the different [14]
patterns and its gains using the statistical the price prediction by Sin et al using
insights. Machine Learning is a subset of genetic algorithms within neural networks.
[15]
Artificial intelligence helps to train the data As per Jang et al. the mean average
[17] percentage error (MAPE) produces a
automatically by improvising algorithms.
regression value of 0.013 using Bayesian
The Sentimental analysis also plays a vital neural networks. Using LTSM neural
role in the predictions of the market. networks with an accuracy of 53% by
Converging of these two technologies, [13]
McNally et al. in a price change.
Blockchain and Machine Learning are
Collective classification models are used in
capable of analysing the data securely and [10]
predicting analysis of graphs
helps the predictive analysis of prices of
cryptocurrency. Problem and Solution:
The Market cap for cryptocurrency as of May Unstructured and bad data in the crypto-
2020 is $265.545 billion and Blockchain currency market provided from the internet
market by 2023 is estimated to reach up to
is mostly manipulated or inaccurate data.
[5]
23.3 billion. Around the globe, 40 million This helps for the fake and bot trading with
users are now using cryptocurrency. In which most of the cryptocurrency exchanges which
bitcoin, the popular known cryptocurrency are most often manipulated. As per the
gained 193,639.36% growth between 2012 bitwise, a report in March 2019, 95% of the
and 2020 with a record of online trades or exchanges are fake also includes
[5]
transactions about $6 billion daily. Coinmarketcap which is one of the popular
[4]
Related works: crypto data websites.

Cryptocurrency is a coin-based market which

The future price change in the prediction in
the market simulation where the digital is volatile in nature compared to the stock
[12]
market. It depends upon the people’s
sentiment regardless to the actual market
strategies. Using supervised learning, which
is one of the Machine Learning algorithms by Methodology:
evaluating predictive analysis of the price
and movement of the cryptocurrencies. This As per the stock market predictions Machine
Learning techniques plays a vital role in
research main contribution is to investigate
generating successful prediction through its
two major questions:
analysis. These Machine Learning techniques
i) Can machine learning algorithms can be considered to predict cryptocurrency
[8]
efficiently predict cryptocurrency prices.
prices?
ii) Any proper validations for the Howbeit, the Machine Learning algorithm is
limited so far in the cryptocurrency market
predictive analysis for the
with the analysis of bitcoin prices. These
cryptocurrency market?
algorithms such as supervised learning and
Machine learning algorithms using random forest will be able to analyse the
[8]
supervised learning, random forest price fluctuations of the bitcoin.
algorithms to analyse a large amount of data
Machine learning models are explicitly
using sentiment analysis. Also, Blockchain
known for pattern recognition and self-
technology is used to secure and store the
learning the methodology which can analyse
data for analysing by supervised machine
the patterns in different ways by evolving
learning algorithm. By the process of
itself to perform appropriate actions to
optimization, another limitation raises by its
propose the best predictions for the
parameters which are to be tested as the
cryptocurrency prices. But this can’t be able
priority of this research is to find out the
to determine in some of the factors such as
predictive memory analysis and historical
psychological factors of human. The truth
memory analysis.
behind the accuracy upon future the market
Cryptocurrency is also criticized because of price is low as it is compared to the historical
its exchange rate volatility, another hand models, psychic models and quantitative
because of its divisibility, resistance, models which are failed for trying with many
[3]
inflation, transparency and portability is also variables of the complex problems.
[6]
praised. The actual prediction of
Using Machine Learning classification
cryptocurrency is to be maintained securely
algorithms along with linear/logistic
with reliable data resources using blockchain
regression, support vendor machine and
technology and machine learning models to
random forests with the comparison of the
give the most accurate report possible to
determine future prices of the past data results in the predictability of the
cryptocurrency market. cryptocurrency prices in the market.
Supervised learning algorithms are one of
the best techniques for predictive analysis. hundred of decision trees to forecast the
Considering the past and present data helps predictions or the best results. Supervised
to estimate future prices. The previous data learning method needs extensive training,
records are used to analyse the market whereas random forest collects the small
capitalization and estimate future prices amount of datasets and provide quick
using Machine learning with the help of evaluations. Cross-correlations and multi-
Blockchain technology. For stock market collinearity doesn’t been considered by
predictions using machine learning random forest we use other methods to
techniques helps to predict the [18]
determine the values.
cryptocurrency prices by using random
forest. A classification approach for the predictive
analysis of cryptocurrency is by random
Opinion mining is also known as sentiment forest algorithm based upon its prediction by
analysis, which is creating an opinion digitally constructing the decision trees. By random
in various social networks. Social media subspace method, it de-correlates the trees
[12]
increase in the present society makes a by changing the bagging algorithm.
much difference in a person’s views upon
using the sentiment analysis. The opinion The prediction for the price changes for each
may be based upon the positive and negative cryptocurrency is approximately showed 8%
[9]
impacts which can affect the trade whereas of accuracy gaps. Few improvements
the cryptocurrency trades might be required for the precision of prediction
depending upon the public views and analysis for the accuracy rate.
opinions.
There are different requirements for each
Supervised learning algorithm in machine machine learning algorithm which always
learning helps to detect the patterns of the looks for the best possible solution for the
data-driven by the cryptocurrency market problem given. As the prediction,
for the predictive analysis and give various cryptocurrency gets an accurate price then
algorithm methods to maintain the the investors are not much worried about
explanatory data variables which are often the volatility of cryptocurrencies. This
mentioned to data variables, statistical assumption can be better and bigger in the
approach and predictors to determine to fit market for the investors to the scope of
[11] profit.
in the prediction model.
Conclusion:
Random forest is one of the machine
learning algorithm method which is based on The main aim of this research is to identify
decision trees. Both regression and the prediction methods where it is used to
classification problems are applied, the investigate whether one can predict the
random forest can implement with a price fluctuations accurately. By this paper,
analysing the data using multiple Machine 3. M. Wimalagunaratne and G. Poravi, "A Predictive
Learning algorithms such as supervised Model for the Global Cryptocurrency Market: A
Holistic Approach to Predicting Cryptocurrency
learning, the random forest along with
Prices," 2018 8th International Conference on
sentiment analysis can be possible in the Intelligent Systems, Modelling and Simulation (ISMS),
determination of predictive analysis of Kuala Lumpur, Malaysia, 2018, pp. 78-83, doi:
cryptocurrency markets. 10.1109/ISMS.2018.00024.
4. CINDY TAYLOR, Bitwise: 95% Of Bitcoin
The machine learning algorithms applied for Trading Volume Is Fake, MARCH 27, 2019
the trading information, data results and 5. Mike Murphy, Cryptocurrencies lose nearly
historic price data can be helped for the $21 billion in total market cap in Sunday selloff,
predictive analysis of cryptocurrencies. The Published: March 9, 2020 at 12:14 a.m. ET
Blockchain technologies are helped to secure 6. Giudici, G., Milne, A. & Vinogradov, D.
the data and more securely and with the use Cryptocurrencies: market analysis and
of machine learning techniques derives the perspectives. J. Ind. Bus. Econ. 47, 1–18 (2020).
https://doi.org/10.1007/s40812-019-00138-6
prediction of cryptocurrencies in the future.
Pintelas E., Livieris I.E., Stavroyiannis S.,
Further Work: Kotsilieris T., Pintelas P. (2020) Investigating the
Problem of Cryptocurrency Price Prediction: A
The implementation of the above research is Deep Learning Approach. In: Maglogiannis I.,
yet to be worked on the given proposed Iliadis L., Pimenidis E. (eds) Artificial Intelligence
solution. This research paves the way for Applications and Innovations. AIAI 2020. IFIP
Advances in Information and Communication
further work, by analysing the algorithm
Technology, vol 584. Springer, Cham.
leads to the improved results. In future,
https://doi.org/10.1007/978-3-030-49186-4_9
additional data can be used for the
7. Pintelas E., Livieris I.E., Stavroyiannis S.,
predictive analysis of cryptocurrency for the
Kotsilieris T., Pintelas P. (2020) Investigating the
precise price rate. Using machine learning
Problem of Cryptocurrency Price Prediction: A
algorithms as given will improve the Deep Learning Approach. In: Maglogiannis I.,
prediction itself by learning the data Iliadis L., Pimenidis E. (eds) Artificial Intelligence
provided. Applications and Innovations. AIAI 2020. IFIP
Advances in Information and Communication
References: Technology, vol 584. Springer, Cham.
https://doi.org/10.1007/978-3-030-49186-4_9
1. Augustin, Peter & Salman, Agha. (2019).
8. Laura Alessandretti, Abeer ElBahrawy, Luca
Predictive Analytics in Cryptocurrency Using
Maria Aiello, Andrea Baronchelli, "Anticipating
Neural networks: A Comparative Study.
Cryptocurrency Prices Using Machine Learning",
International Journal of Advanced Technology
Complexity, vol. 2018, Article ID 8983590, 16
and Engineering Exploration. 7. 425.
2. A. ElBahrawy, L. Alessandretti, A. Kandler, R. Pastor- pages, 2018.
https://doi.org/10.1155/2018/8983590
Satorras and A. Baronchelli, "Evolutionary dynamics of the
cryptocurrency market", Royal Society open science, vol. 4, 9. Young Bin Kim,Jun Gi Kim,Wook Kim,Jae Ho
no. 11, pp. 170623, 2017 Im,Tae Hyeong Kim,Shin Jin Kang,Chang Hun
 Kim, Predicting Fluctuations in Cryptocurrency Distributed and Network-based Processing
Transactions Based on User Comments and (PDP), 2018.
Replies, Published: August 17, 2016 14. E. Sin and L. Wang, "Bitcoin price
https://doi.org/10.1371/journal.pone.0161197 prediction using ensembles of neural
10. X. Kong, P. S. Yu, Y. Ding and D. J. Wild, networks", 2017 13th International
"Meta path-based collective classification in
Conference on Natural Computation Fuzzy
heterogeneous information networks", Systems and Knowledge Discovery (ICNC-
Proceedings of the 21st ACM international FSKD), pp. 666-671, 2017.
conference on Information and knowledge
15. H. Jang and J. Lee, "An empirical study
management, pp. 1567-1571, 2012.
on modeling and prediction of bitcoin prices
with bayesian neural networks based on
11. Giatsidis, I., Kitsios, F. and Kamariotou, M. blockchain information", IEEE Access, vol. 6,
(2019). Digital Transformation and User
pp. 5427-5437, 2018.
Acceptance of Information Technology in the
Banking Industry, Proceedings of the 8th 16. C. Bai, T. White, L. Xiao, V. S.
International Symposium & 30th National Subrahmanian and Z. Zhou, "C2P2: A
Conference on Operational Research, Patras, Collective Cryptocurrency Up/Down Price
Greece, pp. 6-10. [see: Prediction Engine," 2019 IEEE International
http://eeee2019.teiwest.gr/docs/HELORS_2019_ Conference on Blockchain (Blockchain),
proceedings.pdf] Atlanta, GA, USA, 2019, pp. 425-430, doi:
12. S. Rahman, J. N. Hemel, S. Junayed Ahmed 10.1109/Blockchain.2019.00065
Anta, H. A. Muhee and J. Uddin, "Sentiment 17. W. Yiying and Z. Yeze, "Cryptocurrency
Analysis Using R: An Approach to Correlate PriceAnalysiswithArtificial
Cryptocurrency Price Fluctuations with Change in Intelligence," 2019 5th International
User Sentiment Using Machine Learning," 2018 Conference on Information Management
Joint 7th International Conference on (ICIM), Cambridge, United Kingdom, 2019,
Informatics, Electronics & Vision (ICIEV) and 2018
pp. 97-101, doi:
2nd International Conference on Imaging, Vision
10.1109/INFOMAN.2019.8714700.
& Pattern Recognition (icIVPR), Kitakyushu,
Japan, 2018, pp. 492-497, doi: 18. Mudassir, M., Bennbaia, S., Unal, D. et
10.1109/ICIEV.2018.8641075. al. Time-series forecasting of Bitcoin prices using
13. S. McNally, J. Roche and S. Caton, high-dimensional features: a machine learning
approach. Neural Comput & Applic (2020).
"Predicting the price of bitcoin using
https://doi.org/10.1007/s00521-020-05129-6
machine learning", 2018 26th Euromicro
International Conference on Parallel

View publication stats